ACL_in_Practice

仁爱英语八年级下各单元短语归纳集团标准化办公室：[VV986T-J682P28-JP266L8-68PNN]Unit5TopicOne1.lookexcited看起来兴奋，激动2.feelhappy/disappointed感到高兴/失望3.tastedelicious/good尝起来美味4.soundwonderful/sweet/great听起来精彩/甜美/好，不错5.smellterrible闻起来恶心6.gomad/bad发疯/变坏7.turngreen/yellow变成绿色/黄色8.bepopularwithsb.受某人欢迎9.seem(tobe)unhappy似乎不高兴10.seemtodosth.似乎做，好像做11.gotothemovies=gotoseethemovie去看电影12.invite/asksbtodosth邀请某人做某事13.invitesb.to+sp.邀请某人去/参加…14.oneof+最高级+pl.最…之一15.preparesthforsb=preparesb.sth.为某人准备好某事16.preparetodosth.准备做17.saythanks/hello/sorry/goodbyetosb向某人说声谢谢/你好/抱歉/再见18.Whatashame/pity.真遗憾。

19.get/buytheticketto/for买到…的票20.beabletodo=can/coulddo能够做…21.willbeabletodo将会做…22.onthe/one’sway to在去往…的路上23.onthewayhome/here/there在回家/去这儿/那儿的路上24.rightnow/away=atonce立即，马上25.feel/besorryfor(doing)sth为（做）某事感到抱歉/遗憾/难过26.beangrywithsb.对某人生气27.justnow=amomentago刚才（用于过去时）28.carefor=takecareof=lookafter照顾29.becauseof+n.因为…，由于…30.because+句子因为…，由于…31.cheersbup=makesb.Happy使…兴奋起来32.atlast=intheend=finally最后，最终33.attheendof…在…的末尾/尽头34.atfirst起初，开始smilingfaces笑脸35.noisychildren吵闹的孩子们36.lovelysongs活泼的歌曲37.livealone独居be/feellonely感到孤独38.teachsb.todosth.教某人做39.teachoneselfsth.=learnsth.byoneself自学…eintobeing=beborn出现，形成41.fallinto落入，掉入lookfor寻找42.everywhere=hereandthere到处43.bewithahistoryof200years=haveahistoryof200years=have200 yearsofhistory有着两百年的历史eintobeing=beborn形成efrom=befrom来自46.moreandmore+n./原级越来越来…47.be/becomeinterestedin(doing)sth.对…感兴趣48.makepeacewithsb与某人和解49.findawaytodo/ofdoingsth.找到做…的方法50.Itseems/seemedthat+句子。

ERTMS/ETCS – Class 1GSM-R InterfacesClass 1 RequirementsREF : SUBSET-093ISSUE : 2.3.0DATE : 10-Oct-2005Company Technical Approval Management approval ALCATELALSTOMANSALDO SIGNALBOMBARDIERINVENSYS RAILSIEMENS1. M ODIFICATION H ISTORYIssue NumberDateSection Number Modification / Description Author0.1.0 (8-Aug-02) Creation based on subset052LK0.1.1 (8-Aug-02) All Minor editorial changes LK0.1.1ec All englishcheck JH0.2.0 (9-Sep-02) 3., 4.2, 4.1, 6.3, 7.2,8.2 Updated after email discussionLK0.3.0 (24-Oct-02) All Updated after FlorencemeetingLK+TS0.4.0 (14-Nov-02) All Updated after LondonmeetingLK0.5.0 (5-Dec-02) 4.2, 5.6.1, 6.2, 7.1,7.3, 9.2 Updated after Berlin meetingLK0.6.0 (12-Dec-02) 3., 6.3., 10.4.3 Email comments included TS+LK2.0.0 (12-Dec-02) Erroneous versionnumber 2.2.0correctedFinal issue LK2.1.0 (28-March-03)3.1.1.1, 6.3.1.3,7.1.1.1, 8.1.1.1 Update acc. to super group commentsLK2.2.0 (28-March-03) - Final version LK2.2.2.31-03-03 Versionnumberchangedfor release to the usersGroupWLH2.2.3 (12-June-03) All Update after Brussels mtg.and GSM-R Op. grp.commentsLK2.2.4 (26-June-03) editorial Draft release to UsersGroupJH2.2.5 - FormalreleaseJH 2.2.5.1 4.2, 6.2, 6.3, new 6.4 Update after Paris mtg. andGSM-R Op. grp. commentsLK2.2.5.2 Various update after further GSM-ROp grp reviewJH2.2.5.3 cleanversion JH 2.2.5.4 6.4 Updated after further GSM-R Op grp requestRB2.2.6 CleanversionRB2.2.6 revA (31-Jan-05) 4.2, 6.3, 6.4, Annex A Proposal for QoS parametervaluesLK2.2.6 revB (14-Feb-05) 6.3, 6.4, Annex A Updated after QoSmeeting#6 BrusselsLK2.2.6 revC (24-Feb-05) 6.3, 6.4, Annex A,Annex B added Updated during BerlinmeetingLK2.2.6 revD (25-Feb-05) 6.2, 6.3.5, 10.3, 10.5.2 Email comments inserted LK2.2.6 revE (6-Apr-05)3.1., 3.2,4.1,5.1,6.3,10.1, 10.3, 10.5, 10.7 Updated after QoSmeeting#7 BrusselsLK2.2.6 revF (25-Apr-05)3.1,4.1,5.1,6.3, 6.4,10.1, 10.3, 10.5, 10.6,10.7Edinburgh meeting TS+LK2.2.6revG (20-May-05)3.1, 5.1, 6.3, 6.4, 8.2 Changes according toBrussels meetingLK2.2.6revH (1-Sep-05) 4.1, 5.1, 6.3, 6.4, 7.2,10.3, 10.4, 10.5 Comments from SG andEEIGLK2.2.6revI (8-Sep-05) 5.1, 6.3, 6.4, 10.4 Zürich meeting PL+LK 2.3.0 (10-Oct-05) update for issue JH2. T ABLE OF C ONTENTS1.M ODIFICATION H ISTORY (2)2.T ABLE OF C ONTENTS (4)3.R EFERENCES (6)3.1Normative Documents (6)3.2Informative Documents (7)4.T ERMS AND DEFINITIONS (8)4.1Abbreviations (8)4.2Definitions (9)5.G ENERAL (10)5.1Scope of this document (10)5.2Introduction (10)6.E ND-TO-END SERVICE REQUIREMENTS TO GSM-R NETWORKS (12)6.1Data bearer service requirements (12)6.2Additional services (12)6.3Quality of Service requirements (13)6.3.1General (13)6.3.2Connection establishment delay (14)6.3.3Connection establishment error ratio (14)6.3.4Transfer delay (15)6.3.5Connection loss rate (15)6.3.6Transmission interference (15)6.3.7GSM-R network registration delay (16)6.4Summary of QoS requirements (16)7.R EQUIREMENTS TO FIXED NETWORK INTERFACE (17)7.1Foreword (17)7.2Interface definition (17)7.3Communication signalling procedures (17)8.R EQUIREMENTS TO MOBILE NETWORK INTERFACE (18)8.1Foreword (18)8.2Interface definition (18)9.A NNEX A(I NFORMATIVE) TRANSMISSION INTERFERENCE AND RECOVERY (19)9.1General (19)9.2Transmission interference in relation to HDLC (19)10.A NNEX B(INFORMATIVE)J USTIFICATION OF Q O S PARAMETER VALUES (22)10.1General (22)10.2Connection establishment delay (22)10.3Connection establishment error ratio (22)10.4Transfer delay (23)10.5Connection loss rate (23)10.5.1QoS targets (23)10.5.2Conclusions (24)10.6Transmission interference (24)10.7Network registration delay (26)3. R EFERENCESDocuments3.1 Normative3.1.1.1 This document list incorporates by dated or undated references, provisions from otherpublications. These normative references are cited at the appropriate place in the textand the publications are listed hereafter. For dated references, subsequentamendments to or revisions of any of these publications apply to this document onlywhen incorporated in it by amendment or revision. For undated references the latestedition of the publication referred to apply.Reference DateTitleU-SRS 02.02 ERTMS/ETCS Class 1; Subset 026; Unisig SRS, version 2.2.2 Subset 037 07.03 ERTMS/ETCS Class 1; Subset 037; EuroRadio FIS; Class1requirements, version 2.2.5EIRENE FRS 10.03 UIC Project EIRENE; Functional Requirements Specification.Version 6.0, CLA111D003EIRENE SRS 10.03 UIC Project EIRENE; System Requirements Specification.Version 14.0, CLA111D004ETS 300011 1992 ISDN; Primary rate user-network interface; Layer 1 specificationand test principlesETS 300102-1 1990 ISDN; User-network interface layer 3; Specification for basiccall controlETS 300125 1991 ISDN; User-network interface data link layer specificationsGSM04.21 12.00 Rate Adaptation on the MS-BSS Interface, v.8.3.0GSM 07.0711.98 ETSI TS 100916; Digital cellular telecommunications system(Phase 2+); AT command set for GSM Mobile Equipment (ME),GSM TS 07.07 version 6.5.0 Release 1997ITU-T V.24 02.00 List of definitions for interchange circuits between data terminalequipment (DTE) and data circuit-terminating equipment (DCE)ITU-T V.25ter 07/97 Serial asynchronous dialling and controlITU-T V.110 02.00 Support of data terminal equipments (DTEs) with V-series typeinterfaces by an integrated services digital network (ISDN) EuroRadio FFFIS 09.03 UIC ERTMS/GSM-R Unisig; Euroradio Interface Group; RadioTransmission FFFIS for Euroradio; A11T6001; version 12O-2475 09.03 UIC ERTMS/GSM-R Operators Group; ERTMS/GSM-R Qualityof Service Test Specification; O-2475; version 1.0Documents3.2 InformativeTitleReference DateEEIG 04E117 12.04 ETCS/GSM-R Quality of Service - Operational Analysis, v0.q(draft)ERQoS 08.04 GSM-R QoS Impact on EuroRadio and ETCS application,Unisig_ALS_ERQoS, v.0104. T ERMS AND DEFINITIONS4.1 AbbreviationsAT ATtention command setATD AT command DialB channel User channel of ISDNB m channel User channel of GSM PLMN on the air interfaceBRI Basic Rate InterfaceByte 1 start bit + 8 data bits + 1 stop bitDCE Data Circuit EquipmentDCD Data Carrier DetectD channel Control channel of ISDND m channel Control channel of GSM PLMN on the air interfaceDTE Data Terminal EquipmenteMLPP enhanced Multi-Level Precedence and Pre-emptionFIS Functional Interface SpecificationGPRS General Packet Radio Service (a phase 2+ GSM service) GSM-R Global System for Mobile communication/RailwayHDLC High level Data Link ControlISDN Integrated Services Digital NetworkMLPP Multi-Level Precedence and Pre-emption (ISDN service) MOC Mobile Originated CallMS Mobile Station (a GSM entity)Termination/Terminated MT MobileMTC Mobile Terminated CallMTBD Mean Time Between DisturbanceUnitOBU On-BoardPLMN Public Land Mobile NetworkPRI Primary Rate InterfaceQoS Quality of ServicesRBC Radio Block CentreT TI Duration of Transmission Interference periodT REC Duration of Recovery periodUDI Unrestricted Digital4.2 Definitions4.2.1.1 Definitions for the purpose of this specification are inserted in the respective sections.5. G ENERAL5.1 Scope of this document5.1.1.1 The scope of this document is to specify the Radio Communication Systemrequirements to the GSM-R network services (including fixed side access) andinterfaces and also the pre-requisites to be fulfilled by GSM-R networks and ETCSinfrastructures. Presently the requirements for high-speed lines are covered,requirements for conventional lines may be included in future versions of thisdocument.5.1.1.2 The data transmission part of the communication protocols is fully described in theEuroRadio FIS [Subset 037].5.1.1.3 The Radio Transmission FFFIS for EuroRadio [EuroRadio FFFIS] specifies thephysical, electrical and functional details related to the interfaces.5.1.1.4 All requirements apply to GSM-R unless indicated otherwise .5.2 Introduction5.2.1.1 The definition of the GSM services and associated physical and communicationsignalling protocols on the air interface are fully standardised in the specificationsproduced by the ETSI GSM Technical Committee for the public GSM implementationas well as for the GSM-R. Additionally, some railway specific services are alsospecified in the EIRENE SRS. However, in both cases, not all are required for ERTMSclass 1 system definition.5.2.1.2 The following ETSI GSM phases 1/2/2+ services are required:a) Transparent data bearer serviceb) Enhanced multi-level precedence and pre-emption (eMLPP).5.2.1.3 Other ETSI GSM phases 1/2/2+ services are not required for Class 1. These are thefollowing :a) GSM supplementary services:• Call forwardingb) General packet radio service (GPRS)5.2.1.4 Other ETSI GSM phases 1/2/2+ services are not required. Examples of these are thefollowing :a) Non-transparent data bearer serviceb) GSM supplementary services:• Line identification•Call waiting and hold• Multiparty•Closed User Group•Advice of charge• Call Barringc) Short message service point to point or cell broadcastd) Voice broadcast servicee) Voice group call service5.2.1.5 The following EIRENE railway specific service [EIRENE SRS] is required:a) Location dependent addressing5.2.1.6 The following EIRENE specific services [EIRENE SRS] are not required :a) Functional addressingb) Enhanced location dependent addressingc) Calling and connected line presentation of functional identitiesd) Emergency callse) Shunting modef) Multiple driver communications6. E ND-TO-END SERVICE REQUIREMENTS TO GSM-RNETWORKS6.1 Data bearer service requirements6.1.1.1 For the transmission of information between OBU and RBC, the EuroRadio protocoluses the bearer services of a GSM-R network. The service provider makes these databearer services available at defined interfaces.6.1.1.2 The data bearer services are described as data access and transfer in the GSMnetwork from Terminal Equipment (TE) on the mobile side (i.e. OBU) to a networkgateway interworking with Public Switched Telephonic Network (PSTN) or IntegratedServices Digital Network (ISDN) on the fixed side (i.e. RBC).6.1.1.3 The following features and attributes of the required bearer service shall be provided:a) Data transfer in circuit switched modeb) Data transfer allowing multiple rate data streams which are rate-adapted[GSM04.21] and [ITU-T V.110]c) Unrestricted Digital Information (UDI) – only supported through ISDN interworking(no analogue modem in the transmission path)d) Radio channel in full ratee) Transfer of data only (no alternate speech/data)f) Transfer in asynchronous transparent modeg) The required data rates are listed in the following table:Bearer service Requirement24. Asynchronous 2.4 kbps T O25. Asynchronous 4.8 kbps T M26. Asynchronous 9.6 kbps T MT: Transparent; M: Mandatory; O: OptionalTable1 GSM-R bearer servicesservices6.2 Additional6.2.1.1 The following supplementary services shall be provided:a) Enhanced multi-level precedence and pre-emption.b) The selection of a particular mobile network shall be possible on-demand.6.2.1.2 The priority value for command control (safety) shall be assigned to according to[EIRENE FRS §10.2] and [EIRENE SRS §10.2].6.2.1.3 The following railway specific service shall be provided by GSM-R networks:a) Location dependent addressing based on the use of short dialling codes inconjunction with cell dependent routing.6.3 Quality of Service requirements6.3.1 General6.3.1.1 As an end-to-end bearer service is used, a restriction of requirements on the servicequality placed on the air interface is not sufficient.6.3.1.2 End-to-end quality of service has to be considered at the service access points.6.3.1.3 The service access points are:•the service access points to the signalling stack for the establishment or release of a physical connection,•the service access points to the data channel.6.3.1.4 The network shall be able to support transparent train-to-trackside and trackside-to-train data communications at speeds up to 500 km/h e.g. in tunnels, cuttings, onelevated structures, at gradients, on bridges and stations.6.3.1.5 The network shall provide a Quality of Service for ETCS data transfer that is at least asgood as listed below1. The parameters are valid for one end-to-end connection for onetrain running under all operational conditions.6.3.1.6 The required QoS parameters shall not depend on network load.6.3.1.7 These performance figures reflect railway operational targets [EEIG 04E117].6.3.1.8 Note: A justification of the performance figures is given by Annex B.6.3.1.9 QoS requirements are specified independently of the method of measurement (refer to[O-2475] for specification of testing).6.3.1.10 Conventional line quality of service requirements may be included in future versions ofthis document. Also the values may not be applied at all locations and times (e.g.discontinuous radio coverage at some locations).6.3.1.11 Given the performance constraints of GSM-R, pre-conditions may be necessary tomeet the railway operational targets of [EEIG 04E117]. If different operational QoStargets are required, then other pre-conditions on ETCS application may be necessary.1 Early experience suggests that GSM-R performance can be better than these parameters suggest, after network optimisation and tuning.Such a case is not covered by this specification and this aspect of ETCS SystemPerformance becomes the responsibility of whoever specifies different operationaltargets.6.3.2 Connection establishment delay6.3.2.1 Connection establishment delay is defined as:Value of elapsed time between the connection establishment request and theindication of successful connection establishment.6.3.2.2 In case of mobile originated calls, the delay is defined between the request bycommand ATD and indication by the later of the two events response CONNECT ortransition of DCD to ON.6.3.2.3 The connection establishment delay of mobile originated calls shall be <8.5s (95%),≤10s (100%).6.3.2.4 Delays>10s shall be evaluated as connection establishment errors.6.3.2.5 The required connection establishment delay shall not depend on user data rate of theasynchronous bearer service.6.3.2.6 The required connection establishment delay is not valid for location dependentaddressing.6.3.3 Connection establishment error ratio6.3.3.1 The Connection establishment error ratio is defined as:Ratio of the number of unsuccessful connection establishment attempts to the totalnumber of connection establishment attempts.6.3.3.2 “Unsuccessful connection establishment attempt” covers all possible types ofconnection establishment errors caused by end-to-end bearer service.6.3.3.3 Connection establishment delays >10s shall be evaluated as connection establishmenterrors.6.3.3.4 The GSM-R networks should be designed in such a way, that at least two consecutiveconnection establishment attempts will be possible (pre-condition on GSM-Rnetworks), e.g. regarding GSM-R radio coverage related to maximal possible trainspeed.6.3.3.5 If the operational QoS targets of [EEIG 04E117] are wanted, then the ETCSinfrastructure should be designed in such a way, that at least two consecutiveconnection establishment attempts will be possible (Recommended pre-condition forETCS infrastructure).6.3.3.6 The connection establishment error ratio of mobile originated calls shall be <10-2 foreach attempt .6.3.3.7 Note: entry into Level 2 is of particular importance; commonly, a time of 40s may berequired in the case the GSM-R mobile station is already registered with the GSM-Rnetwork (see [ERQoS]).6.3.4 Transfer delay6.3.4.1 The end-to-end transfer delay of a user data block is defined as:Value of elapsed time between the request for transfer of a user data block and theindication of successfully transferred end-to-end user data block6.3.4.2 The delay is defined between the delivery of the first bit of the user data block at theservice access point of transmitting side and the receiving of the last bit of the sameuser data block at the service access point of the receiving side.6.3.4.3 The end-to-end transfer delay of a user data block of 30 bytes shall be ≤0.5s (99%).6.3.5 Connection loss rate6.3.5.1 The Connection loss rate is defined as:Number of connections released unintentionally per accumulated connection time.6.3.5.2 The requirements for connection loss rate varies depending on ETCS system variablessuch as T_NVCONTACT and the possible train reactions after connection loss (seesection 10.5).6.3.5.3 If the operational QoS-targets of [EEIG 04E117] are wanted, then the ETCSinfrastructure should be designed in such a way, that at least the following conditionsare fulfilled (Recommended pre-condition for ETCS infrastructure):• T_NVCONTACT ≥ 41s and• M_NVCONTACT different to train trip and• a new MA reach the OBU before standstill.6.3.5.4 If the connection establishment error ratio is <10-2, then the connection loss rate shallbe <10-2/h.6.3.6 Transmission interference6.3.6.1 A transmission interference period T TI is the period during the data transmission phaseof an existing connection in which, caused by the bearer service, no error-freetransmission of user data units of 30 bytes is possible.6.3.6.2 A transmission interference happens, if the received data units of 30 bytes deviatepartially or completely from the associated transmitted data units.6.3.6.3 The transmission interference period shall be < 0.8s (95%), <1s (99%).6.3.6.4 An error-free period T Rec shall follow every transmission interference period to re-transmit user data units in error (e.g. wrong or lost) and user data units waiting to beserved.6.3.6.5 The error-free period shall be >20s (95%), >7s(99%).6.3.7 GSM-R network registration delay6.3.7.1 The GSM-R network registration delay is defined as:Value of elapsed time from the request for registration to indication of successfulregistration by +CREG response.6.3.7.2 The GSM-R network registration delay shall be ≤30s (95%), ≤35s (99%).6.3.7.3 GSM-R network registration delays > 40 s are evaluated as registration errors.6.4 Summary of QoS requirements6.4.1.1 Table 2 contains the summary of QoS requirements at GSM-R interface.QoS Parameter Value (see 6.3) Connection establishment delay of mobile< 8.5s (95%), ≤10s (100%) originated callsConnection establishment error ratio <10-2≤ 0.5s (99%)Maximum end-to-end transfer delay (of 30 bytedata block)Connection loss rate ≤ 10-2 /hTransmission interference period < 0.8s (95%), <1s (99%)Error-free period >20s (95%), >7s(99%)Network registration delay ≤30s (95%), ≤35s (99%), ≤40s (100%)Table 2 Summary of QoS requirements7. R EQUIREMENTS TO FIXED NETWORK INTERFACE7.1 Foreword7.1.1.1 This part of the specification does not define mandatory requirements forinteroperability. It is a preferred solution, in case interchangeability between tracksideRBC and access point to the fixed network is required for a given implementation.7.1.1.2 This section gives only limited information. [EuroRadio FFFIS] must be used for fullcompliance.7.1.1.3 Note: The requirements to fixed network interface refer to a set of ETSI specifications[ETS 300011, ETS 300125, ETS 300102-1]. This set is the basis of conformancerequirements for network terminations. Instead of these specifications updatedspecifications can be referred, if they state that they are compatible with the followingrequirements.7.2 Interfacedefinition7.2.1.1 The ISDN Primary Rate Interface (PRI) shall be provided as specified by [ETS300011].7.2.1.2 The service access point on the fixed network side corresponds with the S2M interfaceat the T-reference point.7.2.1.3 The Basic Rate interface might also be used as an option in some particular cases likeradio infill unit.7.2.1.4 In addition to these interfaces, the V.110 rate adaptation scheme shall be applied tothe user data channel. The RA2, RA1 and RA0 steps are mandatory.7.2.1.5 End-to-end flow control in layer 1 shall not be used.7.3 Communication signalling procedures7.3.1.1 The signalling protocols shall be provided as specified by:a) Link Access Procedure on the D channel [ETS 300125]b) User-network interface layer 3 using Digital Subscriber Signalling [ETS 300102-1]7.3.1.2 ISDN multi-level precedence and pre-emption (MLPP) supplementary service shall beprovided according to the EIRENE specification [EIRENE SRS].7.3.1.3 The SETUP message contains Information Elements including the bearer capabilityand the low layer compatibility (refer to [EuroRadio FFFIS] specifying the Euroradiodata bearer service requirements.8. R EQUIREMENTS TO MOBILE NETWORK INTERFACE8.1 Foreword8.1.1.1 This part of the specification does not define mandatory requirements forinteroperability. It is a preferred solution, in case interchangeability between OBU andMobile Terminal is required for a given implementation.8.1.1.2 This section gives only limited information. [EuroRadio FFFIS] must be used for fullcompliance.definition8.2 Interface8.2.1.1 If an MT2 interface is used at the mobile side, the service access point at the mobilestation corresponds with the R-reference point of the MT2.8.2.1.2 [GSM 07.07] specifies a profile of AT commands and recommends that this profile beused for controlling Mobile Equipment functions and GSM network services through aTerminal Adapter.8.2.1.3 For the mobile termination type MT2 the signalling over the V interface has to be inaccordance with [GSM 07.07], using the V.25ter command set.8.2.1.4 The online command state shall not be used to guarantee interoperability. To avoiddifferent behaviour, it is recommended to enable/disable this escape sequence usingthe appropriate AT command usually referred as ATS2=<manufacturer defined value>.This particular command shall be sent to the mobile terminal as part of its initialisationstring.8.2.1.5 State control using physical circuits is mandatory.8.2.1.6 The V-interface shall conform to recommendation ITU-T V.24. The signals required arespecified in [EuroRadio FFFIS].8.2.1.7 Note that in the case of class 1 mobile originated calls, it is allowed to set the priorityvalue “command control (safety)” at subscription time.8.2.1.8 The call control commands, interface control commands and responses used on the V-interface at the R reference point are specified in [EuroRadio FFFIS].9. A NNEX A(I NFORMATIVE) TRANSMISSION INTERFERENCEAND RECOVERY9.1 General9.1.1.1 The usual QoS parameter used as measure of accuracy of data transmission viatransparent B/B m channels is the bit error rate.9.1.1.2 The QoS parameter relevant for layer 2 accuracy is the HDLC frame error rate.9.1.1.3 It is not possible to define relationships between both rates. The channel behaviour isnot known: error bursts and interruptions of data transmission during radio cellhandover can happen.9.1.1.4 Additionally, statistical distributions of values such as error rates do not accurately mapthe requirements from the ETCS point of view. Transfer of user data is requested inbursts; the transfer delay can be critical for the application. It has to be guaranteed forsome application messages that data can be transferred to the train in a defined timeinterval.9.1.1.5 A model of service behaviour is necessary reflecting all relevant features of GSM-Rnetworks.9.1.1.6 This model can be used as a normative reference for acceptance tests and for networkmaintenance during ETCS operation. It enables the ETCS supplier to demonstrate thecorrect operation of ETCS constituents during conformance testing without thevariations of real world GSM-R networks.9.1.1.7 Transmission interference and recovery is a first approximation of such a servicebehaviour model.9.2 Transmission interference in relation to HDLC9.2.1.1 Transmission interference is characterised by a period in the received data streamduring which the received data units deviate partially or completely from those of thetransmitted data stream. The service user cannot see the causes of transmissioninterference.9.2.1.2 The user data units erroneously transmitted or omitted during the transmissioninterference must be corrected by re-transmission. These re-transmissions result in atime delay and in higher load in the B/B m channel. Therefore, after transmissioninterference a period of error-free transmission, called the recovery period, must follow.9.2.1.3 In the normal data transfer phase after recovery, user data units are transmitted toprovide the data throughput requested by application messages.9.2.1.4 Figure 1 shows a simplified relationship of B/B m channel and HDLC errors: because ofthe selected options for the HDLC protocol (e.g. multi selective reject) the recoveryperiod and the normal data transfer phase are not strictly separated.error-free frameHDLC statecorrupted frameerror-freeChannel stateerroneousFigure 1 B/B m channel and HDLC errors9.2.1.5 Some special cases exist in Figure 1:A Beginning of HDLC frame (corrupted by transmission) is earlier than beginning oftransmission interferenceB Error-free time is not sufficient for transfer of HDLC frameC No HDLC frame is ready for transferD End of corrupted HDLC frame is later than end of transmission interference9.2.1.6 Figure 2 shows as an example the HDLC behaviour in case of transmissioninterference.Figure 2 Event "Transmission interference"9.2.1.7 The sender does not receive an acknowledgement in the case of a corrupted last Iframe of a sequence of I frames. The timer T1 expires and a RR (poll bit set) frame willbe sent.9.2.1.8 After receiving an RR frame with an indication of successful transmission of thepreceding I frame, the lost I frame will be re-transmitted.9.2.1.9 Again the sender does not receive an acknowledgement and requests for thesequence number. Eventually, the transmission is successful but the delivery of userdata will be delayed towards the receiver.9.2.1.10 The occurrence of the above defined event represents a QoS event “Transmissioninterference” at the sender side. The beginning and the end of the transmissioninterference are not exactly known. But the second repetition clearly indicates an event“Transmission interference”:a) The transmission interference time was too long orb) The recovery time was too short.。

rate-‎l imit‎详解‎1. 在全‎局模式下开‎启cef：‎‎Rout‎e r(co‎n fig)‎#ip c‎e f ‎cef(‎c isco‎expr‎e ss f‎o rwar‎d ing,‎c isco‎特快交换）‎‎2. 定‎义标准或者‎扩展访问列‎表：‎ Ro‎u ter(‎c onfi‎g)#ac‎c ess-‎l ist ‎2 per‎m it 1‎92.16‎8.6.0‎0.0.‎0.255‎‎3. 在‎希望限制的‎端口上进行‎r ate-‎l imit‎：‎ Rou‎n ter(‎c onfi‎g-if)‎#rate‎-limi‎t out‎p ut a‎c cess‎-grou‎p 2 1‎28000‎1600‎0 160‎00 co‎n form‎-acti‎o n tr‎a nsmi‎t exc‎e ed-a‎c tion‎drop‎rat‎e-lim‎i t的命令‎格式：‎ #‎r ate-‎l imit‎{inp‎u t|ou‎t put}‎[acc‎e ss-g‎r oup ‎n umbe‎r] bp‎s bur‎s t-no‎r mal ‎b urst‎-max ‎c onfo‎r m-ac‎t ion ‎a ctio‎n exc‎e ed-a‎c tion‎acti‎o n‎ in‎p ut|o‎u tput‎：这是定义‎数据流量的‎方向。

‎ a‎c cess‎-grou‎p num‎b er：定‎义的访问列‎表的号码。

‎‎bps：‎定义流量速‎率的上限，‎单位是bp‎s。

‎ bu‎r st-n‎o rmal‎burs‎t-max‎：定义的数‎据容量的大‎小，一般采‎用8000‎，1600‎0，320‎00，单位‎是字节，当‎到达的数据‎超过此容量‎时，将触发‎某个动作，‎丢弃或转发‎等，从而达‎到限速的目‎的。

‎ co‎n form‎-acti‎o n和ex‎c eed-‎a ctio‎n：分别指‎在速率限制‎以下的流量‎和超过速率‎限制的流量‎的处理策略‎。

目录1 ACL配置............................................................................................................................................1-11.1 ACL简介............................................................................................................................................1-11.1.1 ACL的分类..............................................................................................................................1-11.1.2 ACL的编号和名称...................................................................................................................1-21.1.3 ACL的匹配顺序.......................................................................................................................1-21.1.4 ACL的步长..............................................................................................................................1-31.1.5 ACL的生效时间段...................................................................................................................1-41.1.6 ACL对IPv4分片报文的处理....................................................................................................1-41.1.7 ACL的应用..............................................................................................................................1-41.2 ACL配置任务简介..............................................................................................................................1-41.3 配置ACL............................................................................................................................................1-51.3.1 配置ACL的生效时间段............................................................................................................1-51.3.2 配置WLAN ACL......................................................................................................................1-51.3.3 配置基本ACL..........................................................................................................................1-61.3.4 配置高级ACL..........................................................................................................................1-71.3.5 配置二层ACL..........................................................................................................................1-91.3.6 配置用户自定义ACL.............................................................................................................1-101.3.7 配置简单ACL........................................................................................................................1-101.3.8 复制ACL...............................................................................................................................1-111.4 ACL显示和维护...............................................................................................................................1-121.5 ACL典型配置举例............................................................................................................................1-121.5.1 IPv4 ACL配置举例................................................................................................................1-121.5.2 IPv6 ACL配置举例................................................................................................................1-131 ACL配置本文将用于IPv4和IPv6的ACL分别简称为IPv4 ACL和IPv6 ACL。

2020,24(23):49-52.实用临床医药杂志Journal of Clinical Medicine in Practice・49・关节镜下全内免打结锚钉和高强度缝线技术修复前交叉韧带胫骨止点撕脱骨折效果王水，丁德刚，戴晓峰，朱余龙(江苏省盐城市射阳县人民医院骨科，江苏盐城，224300)摘要：目的探讨关节镜下全内免打结锚钉和高强度纤维缝线技术修复前交叉韧带(ACL)胫骨止点撕脱骨折的效果。

方法选取18例ACL胫骨止点撕脱性骨折的患者，采用关节镜下全内免打结锚钉和高强度纤维缝线技术治疗。

通过LyshUm 评分、国际膝关节文献委员会膝关节评估表(IKDC)评分、Tegaer评分评估膝关节功能恢复情况。

采用Lactman试验、膝关节屈伸活动范围(ROM)评价膝关节稳定性和活动度。

结果术后随访显示，所有患者骨折均愈合，无术后关节内感染、内固定失效等情况。

Lactman试验结果呈阴性,LyhUm评分、IKDC评分、Tegaer评分、膝关节活动度均显著优于术前,差异有统计学意义(P<0.05)o结论关节镜下全内免打结锚钉和高强度纤维缝线技术修复ACL胫骨止点撕脱骨折效果显著，具有创伤小，固定牢,可早期进行膝关节康复锻炼等优点。

关键词：关节镜；前交叉韧带；撕脱骨折；免打结锚钉；高强度缝线；全内技术中图分类号：R683.04；R687.0文献标志码：A文章编号：1672-2353(2220)23-049-04DOI：10.7619/jcmp.222223016Effect of knotless anchor fixation and high-strengthtechnique fiOse sutrre undse ai-iosiOs arthrescopyio repairing anterioe cruciate liqamcne avulsionfrectere of hbiai ioserhonWANG Shui,DING Depang,DAI Xiaoferg,ZHU Yulong(Department of'Orthopedics,She-yang County People's Hospital in Yancheng Cityof Jiangsu Proviccs,Yancheng,Jiangsp,224300)Abstrect:Objechvv To explore effect of all-iasiCd arthroscouia tectniqrd fun redairina antenun cmciaid liqameci(ACL)nvulsiou fractrrn of tiUiai insertiou with knotlesc ancOorr ant high-stncath fiUcr sctcrec.Methods Eightcc patieclc with ACL avulsion fractcra of tibidi insertion were treated wth knotlesc ancOor fixdtion ant Oiga-strength tecOniqca finer sutcra cnCar all-insina arthroscony-T0a fcnctionl recavera of tO knea joints wnc evalrated by Lysholm score,International Knea Docomeatci tioo Committea(IKDC)score ana Teaaer score.TOr scnility c O ranae of motioo of C c kaea were r-vulnated by Lacoman tesi ani knr flexioo ani extedsion ranar of motioo(ROM).Reselts Postooera-tivu foimw-up s C ow C that all the fractnrac0404withool postooerativa joint infectioo ani faimra of in-temal coman tesi was neaativa,anf Lysiolm score,IKDC score,丁跆1^score anf knec ranfc of motion wera31X111X001107ertci'than thosc befora shraera(P<0.05).Conclusion Knoe lesc dnehor fixation anf aiga-stredath tecOniqpa fiUcr shtpra pnnar all-insiVc dithascopy Oava sivnifi-cant effects in匸^£10110anterior cmciatc0x111^1avulsion fractpra of tifial insertion anf they Oava thc of less tranma ,strora fixation,anf ecrly knec redanilitation exercise anf so on.Key words:arthroscony;anterior cmciate lixamedt;1^10X0000110；knotlesc afcOor；Oigh-stanath fiVar sutare；all-insiVa techniqaa前交叉韧带(ACL)胫骨止点撕脱骨折多见骨止点撕脱骨折亦不少见，其临床发病率占于儿童及青少年[9]。

基金项目：泸州市科技局-西南医科大学应用基础研究（2020LZXNYDJ10；2020LZXNYDJ14）通信作者：张驰，E-mail：*********************引用本文：谢羽婕，季玉秀，张弛.前交叉韧带损伤诊治策略[J].西南医科大学学报，2023，46(6)：483-488.DOI：10.3969/j.issn.2096-3351.2023.06.005专家简介：张驰，西南医科大学康复医学系主任，西南医科大学附属医院康复医学科副主任医师，副教授，硕士研究生导师。

四川省卫生健康委员会学术技术带头人后备人选，中华医学会物理医学与康复学分会第十二届委员会青年学组成员，四川省康复医学会冲击波医学专业委员会副主任委员。

获国家自然科学基金、省市校课题项目十余项，获得中国康复医学会科学技术奖二等奖1项，四川省医学青年科技奖二、三等奖3项。

参编专著1部，任副主编，发表学术论文40余篇。

前交叉韧带(anterior cruciate ligament ，ACL)损伤是一种常见的膝关节损伤，占膝关节韧带损伤的50%以上[1]，多见于运动活跃的青少年人群[2]。

在美国每年约有超过20万人受到影响，治疗成本超过70亿美元[3]。

参加高水平比赛运动活跃的年轻人面临更大的风险，40%的受伤归因于非接触性机制，比如快速扭转、跳跃着地[1]。

一旦ACL 损伤，不仅会造成膝关节短期的疼痛、肿胀、活动功能受限，严重者还会出现膝关节弹响、卡顿、关节不稳定感，导致创伤性骨关节炎[4]。

经康复治疗后仍存在膝关节不稳的患者往往需手术干预治疗[5]。

目前针对ACL 损伤患者的手术治疗指征、术后康复策略以及重返运动等关键问题仍然存在争议。

本文对前交叉韧带重建术(anterior cruciate ligament recon‑struction ，ACLR)后康复治疗进行详细阐述，以期为临床提供参考和借鉴。

1前交叉韧带损伤机制与诊断ACL 损伤多为急性损伤，最常见于运动损伤，其次为交通意外[1]。

PART Ⅰ: ChoiceB 1. Which of the following services does not the transport layer provide for the application layer?A.In-order delivery of data segments between processesB.Best effort delivery of data segments between communicating hostsC.Multiplexing and demultiplexing of transport layer segmentsD.Congestion controlA 2. What are the two of the most important protocols in the Internet?A. TCP and IPB. TCP and UDPC. TCP and SMTPD. ARP and DNSC 3. The Internet provides two services to its distributed applications: a connection oriented reliable service and a ( ).A. connection oriented unreliable serviceB. connectionless reliable serviceC. connectionless unreliable serviceD. In order data transport serviceD 4. Processes on two different end systems communicate with each other by exchanging ( ) across the computer network.A. packetsB. datagramC. framesD. messagesA 5. The job of delivering the data in a transport-layer segment to the correct socket is called ( ).A. demultiplexingB. multiplexingC. TDMD. FDMC 6. Two important reasons that the Internet is organized as a hierarchy of networks for the purposes of routing are:A.Least cost and maximum free circuit availabilityB.Message complexity and speed of convergenceC.Scale and administrative autonomyD.Link cost changes and link failureB 7. Which of characters is not distance-vector algorithm’s characters？（）A. iterativeB. globalC. asynchronousD. distributedD 8. The length of IPV6 address is （）bits.A. 32B. 48C. 64D. 128C 9. The host component of a CIDR address of the form a.b.c.d/25 can contain addresses for:A.225 hosts (minus “special” hosts)B.512 hosts (minus “special” hosts)C.2(32-25) hosts (minus “special” hosts)D.25 hosts (minus “special” hosts)C 10. The primary function of the address resolution protocol (ARP) that resides in Internet hosts androuters is:A.To provide LAN router functionsB.To translate between LAN addresses and physical interface addressesC.To translate IP addresses to LAN addressesD.To calculate the shortest path between two nodes on a LANA 11. The POP3 protocol runs over ____ and uses port ____.A. TCP 110B. UDP 110C. UDP 25D. TCP 25D 12.When a destination host transport layer receives data from the network layer, it unambiguouslyidentifies the appropriate process to pass the data to by using a triplet consisting of:A. Source port #, destination IP address, and source IP addressB. Destination port #, source port #, process ID#C. Destination port #, source port #, destination IP addressD. Destination port #, source port #, source IP addressD 13. From the list below, select the items found in the TCP segment structure that are not found in theUDP segment structure:A. Application Generated DataB. Destination Port #C. Source Port #D. Sequence #A 14. The RIP routing protocol is based on an algorithm that is:A. Based on information received only from link “neighbors”B. A link state algorithmC. An OSPF algorithmD. A centralized routing algorithmB 15. With an exterior routing protocol, which of the following issues generally dominates the routing decisions?A. Geographical distance between AS’sB. PolicyC. Number of AS’s traversedD. Current congestion levels in the AS’sA 1. End system are connected together by ____.A. communication linksB. application layerC. transport layerD. the network layerC 2. Which application’s NOT using TCP?A. SMTPB. HTTPC. DNSD. All of themB 3. In the polling protocols, the master node polls each of the nodes in a/an ____ fashion.A. randomB. appointedC. round-robinD. uncirculatedC 4. The DNS protocol runs over ____ and uses port ____.A. UDP 36B. TCP 36C. UDP 53D. TCP 53A 5. TCP provides a ____ service to its applications to eliminate the possibility of the sender over-flowingthe receiver’s buffer.A. flow-controlB. congestion controlC. reliability controlD. data connectionD 6. We can classify just about any multiple access protocol as belonging to one of three categories: channel partitioning protocols, random access protocols, and ____.A. address resolution protocolsB. Dynamic host configuration protocolsC. link-control protocolsD. taking-turns protocolsB 8. The maximum transfer unit(MTU) in Ethernet frame structure is ()byte .A. 1000B. 1500C. 800D. 2000B 9. The socket of UDP is identified by _____ and _______.A. source IP address and source port numberB. destination IP address and destination port number.C. source IP address and destination port number.D. destination IP address and source IP address.C 10. Which is not plug and play in the following four items?A. DHCPB. HubsC. RoutersD. SwitchesD 11．Which of routers is not default routers ?A. first-hop routerB. source routerC. destination routerD. second-hop routerB 13. ICMP is_____.A. the protocol of Application layerB. the protocol of network layerC. the protocol of transport layerD. not a part of TCP/IP protocolsB 14. As general, we has following channel partitioning protocols except ____.A. TDMB. CSMAC. FDMD.CDMAD 15. ____ is most used for error reporting.A. UDPB. SMTPC. FTPD. ICMPB 16. The header of IPV6 is ____byte.A. 20B. 40C. 60D. 80B 17. In the network layer these service are host-to-host service provided by ____. （B）A. the transport layer to the network layerB. the network layer to the transport layerC. the network layer to the network layerD. the transport layer to the transport layerA 18. If there is not enough memory to buffer an incoming packet , a policy that drop the arriving packet called ____.A. drop-tailB. packet lossC. protocolD. encapsulationC 19. In either case, a ____ receives routing protocol messages, which are used to configure its forwarding table.A. serverB. hostC. routerD. ModemD 20. Which of the following functions does not belong to PPP___.A. framingB. link-control protocolsC. network-control protocolsD. error correctionB 1. Which of the following services does the Internet network layer provide for the Internet transport layer?A.In-order delivery of data segments between processesB.Best effort delivery of data segments between communicating hostsC.Multiplexing and demultiplexing of transport layer segmentsD.Congestion controlD 2. The main task of the Internet’s Domain Name System (DNS) is to:A.Translate port numbers to IP addressesB.Specify the standards for Internet domain namesC.Provide an authority for registering domain namesD.Translate mnemonic（记忆的）names to IP addressesA 10. The FTP protocol runs over ____ and uses port ____.A. TCP 21B. TCP 80C. UDP 20D. TCP 110C 3.RDT3.0’s receiver FSM is same to:a) RDT1.0 b) RDT2.1 c) RDT2.2 d) RDT2.0B 4.The Transmission Control Protocol (TCP) provides which of the following services?a)End-to-end station addressingb)Application multiplexingc)Inter network routingd)Medium access control (MAC)D 6.Given that the requested information is not available at any intermediate databases, a non-iterated DNS query from a requesting host would follow the path:a)Root name server, local name server, authoritative name serverb)Authoritative name server, root name server, host name serverc)Local name server, root name server, local name server, authoritative name servere)Local name server, root name server, authoritative name serverA 8.lect the four essential steps, briefly described, for terminating a TCP connection between a client and a server, assuming that the initiating host is the client:(1)Client sends TCP segment with ACK0 and final sequence number(2)Client sends TCP segment with FIN =1 and goes into FIN_WAIT state(3)Server sends TCP segment to ACK the client’s FIN request and enters CLOSE_WAIT state(4)Server sends TCP segment with FIN=0(5)Server sends TCP segment with FIN=1(6)Client sends TCP segment with to ACK server’s FIN and enters second FIN_WAIT state(7)Client sends TCP segment with FIN=0a) 2，3，5，6 b) 5，1，2，3 c) 1，3，5，7 d) 2，3，4，6B 10.When compensating for link cost changes in the distance vector algorithm, it can generally be said that:a)Increased costs are propagated quickly, i.e., “bad news” travels fastb)Decreased costs are propagated rapidly, i.e., “good news” travels fastc)Decreased costs do not converged)None of the aboveB 14.As an IP datagram travels from its source to its destination:a)the source IP address is changed at each router to identify the sending routerb)the router uses the destination IP address to consult its routing tablec)the router does not use the IP addresses in the datagramd)the destination IP address is changed at each router to reflect the next hopC 15.From the list below, choose the bit pattern which could be a valid generator value for the CRC code (R) 11010:a)1110b)011010c)100101d)10011A 16.Consider sending a 1300 byte IPv4 datagram into a link that has an MTU of 500 bytes:a)Three fragments are created.b)Four fragments are created.c)Three fragments are created with offsets 0, 500 1000d)The last fragment consists of exactly 300 bytes of data from the original datagramC 17.Suppose one IPv6 router wants to send a datagram to another IPv6 router, but the two are connected together via an intervening IPv4 router. If the two routers use tunneling, then:a)The sending IPv6 router creates an IPv4 datagram and puts it in the data field of an IPv6datagram.b)The sending IPv6 router creates one or more IPv6 fragments, none of which is larger than themaximum size of an IPv4 datagram.c)The sending IPv6 router creates an IPv6 datagram and puts it in the data field of an IPv4datagram.d)The sending IPv6 router creates an IPv6 datagram and intervening IPv4 router will reject theIPv6 datagramD 18.Which of the following was an important consideration in the design of IPv6a)fixed length 40-byte header and specified options to decrease processing time at IPv6 nodesb)128-bit addresses to extend the address spacec)different types of service (flows) definedd)all of the aboveD 19.A network bridge table is used to perform the following:a)Mapping MAC addresses to bridge port numbersb)Forwarding frames directly to outbound ports for MAC addresses it handlesc)Filtering (discarding) frames that are not destined for MAC addresses it handlesd)All of the abovePART Ⅱ: True / False (1 points per question – total:20 points)1. The DNS server can update the records. (T)2. The TCP connection is a direct virtual pipe between the client’s socket and the server’s connection socket. （T）3. SMTP protocol connect the sender’s mail server and receiver’s mail server (T)4. Whereas a transport-layer protocol provides logical communication between processes running on different hosts, a network-layer protocol provides logical communication between hosts. (T)5. UDP and TCP also provide integrity checking by including right-detection fields in their headers. (F)6. If the application developer chooses UDP instead of TCP, then the application is not directly talking with IP. ( F )7. When we develop a new application, we must assign the application a port number. ( T )8. Real-tine applications, like Internet phone and video conferencing, react very poorly to TCP’s congestion control. ( T )9. The sender knows that a received ACK or NAK packet was generated in response to its most recently transmitted data packet. (T)10. To simplify terminology, when in an Internet context, we refer to the 4-PDU as a unit. (F)11. DV algorithm is essentially the only routing algorithm used in practice today in the Internet。

A Patient’s Guide toA PPatellar Tendon Graft Reconstruction of the ACLMontana Spine & Pain Center500 W. Broadway3rd FloorMissoula, MT 59802Phone: 406-327-1670 Fax: 406-329-5697All materials within these pages are the sole property of Medical Multimedia Group, LLC and are used herein by permission. eOrthopod is aregistered trademark of Medical Multimedia Group, LLC.Montana Spine & Pain CenterPlease take the time to explore our web office. Discover all we have to offer. We hope you willfind the time spent on our website rewarding and informative. Here at Montana Spine & Pain Center, we are dedicated to providing ways for those we serve to access the information needed to make informed decisions about healthcare in orthopaedic and sports medicine. We encourage you to explore our site and learn more about our practice, staff, facilities and treatment options. Check out the Patient Resources section of our site. You will find educational materials to help you understand orthopaedic problems and what options for treatment are available in our clinic.The Staff of the Montana Spine & Pain Center.Montana Spine & Pain Center 500 W. Broadway 3rd FloorMissoula, MT 59802Phone: 406-327-1670 Fax: 406-329-5697IntroductionThe (ACL) is a(Aanterior cruciate ligamentmajor stabilizer of the knee joint. This key knee ligament is commonly torn during sports activities. The standard operation to fix a torn ACL is with a patellar tendon graft. The surgeon takes out the middle section of the patellar tendon below thepatellapatkneecap (). This new graft includes the strip of tendon, along with attached plugs of bone on each end. For this reason, it is sometimes referred to as a bone-patellar-tendon-bone graft. The surgeon removes the torn ACL and puts the new graft into the knee, making sure to line it up just like the original ligament.Many types of tissue grafts have been tried. The patellar tendon graft has proven to be one of the strongest for ACL reconstruction. Patients who have this operation generally get back to their usual activities and sports. They often do so faster than people who have their ACL reconstructed with other types of tissue grafts.This guide will help you understand• what parts of the knee are treated during surgery• how surgeons perform the operation • what to expect before and after the procedure AnatomyWhat parts of the knee are involved? Ligaments are tough bands of tissue that connect the ends of bones together. The ACL is located in the center of the knee joint where it runs from the backside of the femur (thigh-bone) to the front of the tibia (shinbone).The ACL runs through a special notch in the femur called the intercondylar notch and attaches to a special area of the tibia called the tibial spine.The patellar tendon is a thick and strong band of connective tissue on the front of the knee. It starts at the bottom of the patella and fastens just below the knee to a bony bump onthe front of the tibia, called the tibial tubercle . When using the patellar tendon as an ACL graft, surgeons remove a strip from the middle of it. The graft includes the bony attachments from the bottom of the patella and from the tibial tubercle.RationaleWhat does the surgeon hope to accomplish?The main goal of ACL surgery is to keep the tibia from moving too far forward under the femur bone and to get the knee functioning normally again.Many surgeons prefer to use the patellar tendon when reconstructing the ACL. The graft is often chosen because it is one of the strongest ACL grafts. It’s easy to get to, holds well in its location, and generally heals fast.The anatomy of the graft helps to speedhealing and to create a solid connection. When the surgeon implants the new graft, the bony plugs on each end of the graft fit inside a tunnel of bone. This means there is bone-to-bone contact. The body treats the contact of these two bony surfaces as it would a broken bone. It responds by healing the two surfaces together. Healing at the bone-to-bone surface fixes the patellar tendon graft in place.PreparationWhat do I need to know before surgery?You and your surgeon should make the decision to proceed with surgery together. You need to understand as much about the procedure as possible. If you have concerns or questions, you should talk to your surgeon.Once you decide on surgery, you need to take several steps. Your surgeon may suggest a complete physical examination by your regular doctor. This exam helps ensure that you are in the best possible condition to undergo the operation.You may also need to spend time with the physical therapist who will be managing your rehabilitation after surgery. This allows you to get a head start on your recovery. One purpose of this preoperative visit is to record a baseline of information. Your therapist will check your current pain levels, your ability to do your activities, and the movement and strength of each knee.A second purpose of the preoperative visit is to prepare you for surgery. Your therapist will teach you how to walk safely using crutches or a walker. And you’ll begin learning some of the exercises you’ll use during your recovery.On the day of your surgery, you will probably be admitted to the surgery center early in themorning. You shouldn’t eat or drink anythingafter midnight the night before.Surgical ProcedureWhat happens during the operation?Most surgeons perform this surgery using anarthroscope, a small fiber-optic TV camerathat is used to see and operate inside thejoint. Only small incisions are needed duringarthroscopy for this procedure. The operationdoesn’t require the surgeon to open the kneejoint.Before surgery you will be placed undereither general anesthesia or a type of spinalanesthesia. The surgeon begins the operationby making two small openings into the knee,called portals. These portals are where thearthroscope and surgical tools are placed intothe knee. Care is taken to protect the nearbynerves and blood vessels.A small incision is also made below thepatella. Working through this incision, thesurgeon takes out the middle section of thepatellar tendon, along with the bone attach-ments on each end. The bone plugs arerounded and smoothed. Holes are drilledin each bone plug to place sutures (strongstitches) that will pull the graft into place.Next, the surgeon prepares the knee toplace the graft. The remnants of the originalligament are removed. The intercondylar notch(mentioned earlier) is enlarged so that nothingwill rub on the graft. This part of the surgery isreferred to as a notchplasty.Once this is done, holes are drilled in the tibiaand the femur to place the graft. These holesare placed so that the graft will run betweenthe tibia and femur in the same direction as theoriginal ACL.The graft is then pulled into position usingsutures placed through the drill holes. Screwsare used to hold the bone plugs in the drillholes.To keep fluid from building up in your knee, the surgeon may place a tube in your knee joint. The portals and skin incision are then stitched together, completing the surgery. ComplicationsWhat problems can happen with this surgery? As with all major surgical procedures, compli-cations can occur. This document doesn’t provide a complete list of the possible compli-cations, but it does highlight some of the most common problems. Some of the most common complications following patellar tendon graft reconstruction of the ACL are• anesthesia complications• thrombophlebitis• infection• problems with the graft• problems at the donor siteAnesthesia ComplicationsMost surgical procedures require that some type of anesthesia be done before surgery. A very small number of patients have problems with anesthesia. These problems can be reac-tions to the drugs used, problems related to other medical complications, and problems due to the anesthesia. Be sure to discuss the risks and your concerns with your anesthesiologist. Thrombophlebitis (Blood Clots) Thrombophlebitis, sometimes called deep venous thrombosis (DVT), can occur after any operation, but is more likely to occur following surgery on the hip, pelvis, or knee. DVT occurs when blood clots form in the large veins of the leg. This may cause the leg to swell and become warm to the touch and painful. If the blood clots in the veins break apart, they can travel to the lung, where they lodge in the capillaries and cut off the blood supply to a portion of the lung. This is called a pulmonary embolism. (Pulmonary means lung, and embolism refers to a fragment of some-thing traveling through the vascular system.) Most surgeons take preventing DVT very seri-ously. There are many ways to reduce the risk of DVT, but probably the most effective is getting you moving as soon as possible after surgery. Two other commonly used preventa-tive measures include• pressure stockings to keep the blood in the legs moving• medications that thin the blood and prevent blood clots from formingInfectionFollowing surgery, it is possible that the surgical incision can become infected. This will require antibiotics and possibly another surgical procedure to drain the infection.Problems with the GraftAfter surgery, the body attempts to developa network of blood vessels in the new graft. This process, called revascularization, takes about 12 weeks. The graft is weakest during this time, which means it has a greater chance of stretching or rupturing. A stretched or torn graft can occur if you push yourself too hard during this period of recovery. When revas-cularization is complete, strength in the graft gradually builds. A second surgery may be needed to replace the graft if it is stretched or torn.Problems at the Donor SiteProblems can occur at the donor site (the area below the patella where the graft was taken from the knee). A major drawback of taking out a piece of the patellar tendon to reconstruct the ACL is that most patients end up having difficulty kneeling down long after surgery. Lingering pain in the front of the knee is also common.A portion of bone is taken from the bottom of the patella during the graft procedure. This can weaken the patella. In rare cases, heavy useof the quadriceps muscle (on the front of the thigh) can cause the patella to fracture. This often requires a second surgery to repair the broken patella.Taking tissue from the center of the patellar tendon can also cause problems. The body attempts to heal the area but sometimes produces too much scar tissue. The extra scar tissue that forms around the donor site may prevent normal motion in the knee. The patellar tendon is not as strong as it was before surgery. In rare cases, this has been linked to a tear in the patellar tendon. Also, the patellar tendon may become easily inflamed. And problems in this area can keep the quadriceps from regaining normal control and strength.After SurgeryWhat should I expect after surgery?You may use a continuous passive motion (CPM) machine immediately afterward to help the knee begin moving and to alleviate joint stiffness. The machine straps to the leg and continuously bends and straightens the joint. This continuous motion is thought to reduce stiffness, ease pain, and keep extra scar tissue from forming inside the joint. The CPM is often used with a form of cold treatment that circulates cold water through hoses and pads around your knee.Most ACL surgeries are now done on an outpatient basis. Many patients go home the same day as the surgery. Some patients stay one to two nights in the hospital if necessary. The tube placed in your knee at the end of the surgery is usually removed after 24 hours. Your surgeon may also have you wear a protective knee brace for a few weeks after surgery. You’ll use crutches for two to four weeks in order to keep your knee safe, but you’ll probably be allowed to put a comfort-able amount of weight down while you’re up and walking.RehabilitationWhat will my recovery be like?Patients usually take part in formal physical therapy after ACL reconstruction. The first few physical therapy treatments are designed to help control the pain and swelling from the surgery. The goal is to help you regain full knee extension as soon as possible.The physical therapist will choose treatments to get the quadriceps muscles toned and active again. Muscle stimulation and biofeedback, which involve placing electrodes over the quadriceps muscle, may be needed at first to get the muscle going again and to help retrain it.As the rehabilitation program evolves, more challenging exercises are chosen to safely advance the knee’s strength and function. Specialized balance exercises are used tohelp the muscles respond quickly and without thinking. This part of treatment is called neuromuscular training. If you need to stop suddenly, your muscles must react with just the right amount of speed, control, and direc-tion. After ACL surgery, this ability doesn’t come back completely without exercise. Neuromuscular training includes exercises to improve balance, joint control, muscle strength and power, and agility. Agility makes it possible to change directions quickly, go faster or slower, and improve starting and stopping. These are important skills for walking, running, and jumping, and especially for sports perfor-mance.When you get full knee movement, your knee isn’t swelling, and your strength and muscle control are improving, you’ll be able to gradu-ally go back to your work and sport activities. Some surgeons prescribe a functional brace for athletes who intend to return quickly to their sports.Ideally, you’ll be able to resume your previous lifestyle activities. However, athletes are usually advised to wait at least six months before returning to their sports. Most patients are encouraged to modify their activity choices. You will probably be involved in a progressive rehabilitation program for four to six months after surgery to ensure the best result from your ACL reconstruction. In the first six weeks following surgery, expect to see the physical therapist two to three times a week. If your surgery and rehabilitation go as planned, you may only need to do a home program and see your therapist every few weeks over the four to six month period.Notes。

acl aclruntime python推理在Python中，可以使用ACL（Adaptive Concurrency Control）库来进行推理。

ACL提供了一个ACLRuntime类，可以用于加载和运行模型。

首先，确保已经安装了acl和acl-runtime模块。

可以使用以下命令来安装：```pip install acl acl-runtime```然后，你需要将模型转换为ACL格式。

可以使用ACL提供的Model Converter工具来完成此操作。

具体的转换方式因模型而异，可以参考ACL的官方文档或示例代码。

一旦模型被转换为ACL格式，并且已经安装了正确的依赖项，你就可以使用ACLRuntime模块来进行推理。

以下是一个简单的示例代码：```pythonimport aclimport acl.runtime as runtimedef main():# 初始化ACLacl.init()# 创建ACL运行时对象runtime = acl.rt.AclRuntime()# 加载模型model_path = 'model.om'model_id = runtime.load_model(model_path)# 创建输入、输出张量input_desc = acl.mdl.get_input_desc_by_index(model_id, 0) output_desc = acl.mdl.get_output_desc_by_index(model_id, 0) input_tensor = acl.rt.create_tensor(input_desc)output_tensor = acl.rt.create_tensor(output_desc)# 设置输入数据input_data = [1, 2, 3, 4] # 示例输入数据acl.rt.set_input_tensor_data(input_tensor, input_data)# 执行推理runtime.run_model(model_id, [input_tensor], [output_tensor])# 获取输出数据output_data = acl.rt.get_output_tensor_data(output_tensor)print(output_data)# 释放资源acl.rt.destroy_tensor(input_tensor)acl.rt.destroy_tensor(output_tensor)acl.mdl.unload(model_id)acl.rt.reset()acl.finalize()if __name__ == '__main__':main()```在以上示例中，首先初始化ACL库，并创建ACL运行时对象。

A Study of Cross-Validation and Bootstrap for Accuracy E stimation and Model SelectionRon KohaviComputer Science DepartmentStanford UniversityStanford, CA 94305ronnykGCS Stanford E D Uh t t p //r o b o t i c s Stanford edu/"ronnykA b s t r a c tWe review accuracy estimation methods andcompare the two most common methods cross-validation and bootstrap Recent experimen-tal results on artificial data and theoretical recults m restricted settings have shown that forselecting a good classifier from a set of classi-fiers (model selection), ten-fold cross-validationmay be better than the more expensive ka\pone-out cross-validation We report on a large-scale experiment—over half a million runs ofC4 5 and aNaive-Bayes algorithm—loestimalethe effects of different parameters on these algonthms on real-world datascts For cross-validation we vary the number of folds andwhether the folds arc stratified or not, for boot-strap, we vary the number of bootstrap sam-ples Our results indicate that for real-worddatasets similar to ours, The best method lo usefor model selection is ten fold stratified crossvalidation even if computation power allowsusing more folds1 I n t r o d u c t i o nIt can not be emphasized enough that no claimwhatsoever 11 being made in this paper that altalgorithms a re equiva lent in practice in the rea l world In pa rticula r no cla im is being ma de tha t ont should not use cross va lida tion in the real world— Wolpcrt (1994a.) Estimating the accuracy of a classifier induced by su-pervised learning algorithms is important not only to predict its future prediction accuracy, but also for choos-ing a classifier from a given set (model selection), or combining classifiers (Wolpert 1992) For estimating the final accuracy of a classifier, we would like an estimation method with low bias and low variance To choose a classifier or to combine classifiers, the absolute accura-cies are less important and we are willing to trade off biasA longer version of the paper can be retrieved by anony mous ftp to starry Htanford edu pub/ronnyk/accEst-long ps for low variance, assuming the bias affects all classifiers similarly (e g esLimates are ")% pessimistic)In this paper we explain some of the assumptions madeby Ihe different estimation methods and present con-crete examples where each method fails While it is known that no accuracy estimation can be corrert allthe time (Wolpert 1994b Schaffer 1994j we are inter ested in identifying a method that ib well suited for the biases and tn rids in typical real world datasetsRecent results both theoretical and experimental, have shown that it is no! alwa>s the case that increas-ing the computational cost is beneficial especiallhy if the relative accuracies are more important than the exact values For example leave-one-out is almost unbiased,but it has high variance leading to unreliable estimates (Efron 1981) l o r linear models using leave-one-out cross-validation for model selection is asymptotically in consistent in the sense that the probability of selectingthe model with the best predictive power does not con-verge to one as the lolal number of observations ap-proaches infinity (Zhang 1992, Shao 1993)This paper \s organized AS follows Section 2 describesthe common accuracy estimation methods and ways of computing confidence bounds that hold under some as-sumptions Section 3 discusses related work comparing cross-validation variants and bootstrap variants Sec lion 4 discusses methodology underlying our experimentThe results of the experiments are given Section 5 with a discussion of important observations We conelude witha summary in Section 62 Methods for Accuracy E s t i m a t i o nA classifier is a function that maps an unlabelled in-stance to a label using internal data structures An i n-ducer or an induction algorithm builds a classifier froma given dataset CART and C 4 5 (Brennan, Friedman Olshen &. Stone 1984, Quinlan 1993) are decision tree in-ducers that build decision tree classifiers In this paperwe are not interested in the specific method for inducing classifiers, but assume access to a dataset and an inducerof interestLet V be the space of unlabelled instances and y theKOHAVI 1137set of possible labels be the space of labelled instances and ,i n ) be a dataset (possibly a multiset) consisting of n labelled instances, where A classifier C maps an unla-beled instance ' 10 a l a b e l a n d an inducer maps a given dataset D into a classifier CThe notationwill denote the label assigned to an unlabelled in-stance v by the classifier built, by inducer X on dataset D tWe assume that there exists adistribution on the set of labelled instances and that our dataset consists of 1 1 d (independently and identically distributed) instances We consider equal misclassifica-lion costs using a 0/1 loss function, but the accuracy estimation methods can easily be extended to other loss functionsThe accuracy of a classifier C is the probability ofcorrectly clasaifying a randoml\ selected instance, i efor a randomly selected instancewhere the probability distribution over theinstance space 15 the same as the distribution that was used to select instances for the inducers training set Given a finite dataset we would like to custimate the fu-ture performance of a classifier induced by the given in-ducer and dataset A single accuracy estimate is usually meaningless without a confidence interval, thus we will consider how to approximate such an interval when pos-sible In order to identify weaknesses, we also attempt o identify cases where the estimates fail2 1 Holdout The holdout method sometimes called test sample esti-mation partitions the data into two mutually exclusivesubsets called a training set and a test set or holdout setIt is Lommon to designate 2/ 3 of the data as the trainingset and the remaining 1/3 as the test set The trainingset is given to the inducer, and the induced classifier istested on the test set Formally, let , the holdout set,be a subset of D of size h, and let Theholdout estimated accuracy is defined aswhere otherwise Assummg that the inducer s accuracy increases as more instances are seen, the holdout method is a pessimistic estimator because only a portion of the data is given to the inducer for training The more instances we leave for the test set, the higher the bias of our estimate however, fewer test set instances means that the confidence interval for the accuracy will be wider as shown belowEach test instance can be viewed as a Bernoulli trialcorrect or incorrect prediction Let S be the numberof correct classifications on the test set, then s is dis-tributed bmomially (sum of Bernoulli trials) For rea-sonably large holdout sets, the distribution of S/h is ap-proximately normal with mean ace (the true accuracy of the classifier) and a variance of ace * (1 — acc)hi Thus, by De Moivre-Laplace limit theorem, we havewhere z is the quanl lie point of the standard normal distribution To get a IOO7 percent confidence interval, one determines 2 and inverts the inequalities Inversion of the inequalities leads to a quadratic equation in ace, the roots of which are the low and high confidence pointsThe above equation is not conditioned on the dataset D , if more information is available about the probability of the given dataset it must be taken into accountThe holdout estimate is a random number that de-pends on the division into a training set and a test set In r a n d o m sub s a m p l i n g the holdout method is re-peated k times and the eslimated accuracy is derived by averaging the runs Th( slandard deviation can be estimated as the standard dewation of the accuracy es-timations from each holdout runThe mam assumption that is violated in random sub-sampling is the independence of instances m the test set from those in the training set If the training and testset are formed by a split of an original dalaset, thenan over-represented class in one subset will be a under represented in the other To demonstrate the issue we simulated a 2/3, 1 /3 split of Fisher's famous ins dataset and used a majority inducer that builds a classifier pre dieting the prevalent class in the training set The iris dataset describes ins plants using four continuous fea-tures, and the task is to classify each instance (an ins) as Ins Setosa Ins Versicolour or Ins Virginica For each class label there are exactly one third of the instances with that label (50 instances of each class from a to-tal of 150 instances) thus we expect 33 3% prediction accuracy However, because the test set will always con-tain less than 1/3 of the instances of the class that wasprevalent in the training set, the accuracy predicted by the holdout method is 21 68% with a standard deviation of 0 13% (estimated by averaging 500 holdouts) In practice, the dataset size is always finite, and usu-ally smaller than we would like it to be The holdout method makes inefficient use of the data a third of dataset is not used for training the inducer 2 2 Cross-Validation, Leave-one-out, and Stratification In fc-fold cross-validation, sometimes called rotation esti-mation, the dataset V is randomly split into k mutuallyexclusive subsets (the folds) , of approx-imately equal size The inducer is trained and tested1138 LEARNINGThe cross-validation estimate is a random number that depends on the division into folds C o m p l e t ec r o s s -v a l id a t i o n is the average of all possibil ities for choosing m/k instances out of m, but it is usually too expensive Exrept for leave-one-one (rc-fold cross-validation), which is always complete, fc-foM cross-validation is estimating complete K-foId cross-validationusing a single split of the data into the folds Repeat-ing cross-validation multiple limes using different spillsinto folds provides a better M onte C arlo estimate to 1 hecomplele cross-validation at an added cost In s t r a t i -fied c r o s s -v a l i d a t i o n the folds are stratified so thaitlicy contain approximately the same proportions of la-bels as the original dataset An inducer is stable for a given dataset and a set of perturbal ions if it induces classifiers thai make the same predictions when it is given the perturbed datasets P r o p o s i t i o n 1 (V a r i a n c e in A>fold C V )Given a dataset and an inducer If the inductr isstable under the pei tur bations caused by deleting theinstances f o r thr folds in k fold cross-validatwn thecross validation < stnnate will be unbiastd and the t a i lance of the estimated accuracy will be approximatelyaccrv (1—)/n when n is the number of instancesin the datasi t Proof If we assume that the k classifiers produced makethe same predictions, then the estimated accuracy has a binomial distribution with n trials and probabihly of success equal to (he accuracy of the classifier | For large enough n a confidence interval may be com-puted using Equation 3 with h equal to n, the number of instancesIn reality a complex inducer is unlikely to be stable for large perturbations unless it has reached its maximal learning capacity We expect the perturbations induced by leave-one-out to be small and therefore the classifier should be very stable As we increase the size of the perturbations, stability is less likely to hold we expect stability to hold more in 20-fold cross-validation than in 10-fold cross-validation and both should be more stable than holdout of 1/3 The proposition does not apply to the resubstitution estimate because it requires the in-ducer to be stable when no instances are given in the datasetThe above proposition helps, understand one possible assumption that is made when using cross-validation if an inducer is unstable for a particular dataset under a set of perturbations introduced by cross-validation, the ac-curacy estimate is likely to be unreliable If the inducer is almost stable on a given dataset, we should expect a reliable estimate The next corollary takes the idea slightly further and shows a result that we have observed empirically there is almost no change in the variance of the cross validation estimate when the number of folds is variedC o r o l l a r y 2 (Variance m cross-validation)Given a dataset and an inductr If the inducer is sta-ble undfi the }>tituibuhoris (aused by deleting the test instances foi the folds in k-fold cross-validation for var-ious valuts of k then tht vartanct of the estimates will be the sameProof The variance of A-fold cross-validation in Propo-sition 1 does not depend on k |While some inducers are liktly to be inherently more stable the following example shows that one must also take into account the dalaset and the actual perturba (ions E x a m p l e 1 (Failure of leave-one-out)lusher s ins dataset contains 50 instances of each class leading one to expect that a majority indu<er should have acruraov about j \% However the eombmation ofthis dataset with a majority inducer is unstable for thesmall perturbations performed by leave-one-out Whenan instance is deleted from the dalaset, its label is a mi-nority in the training set, thus the majority inducer pre-dicts one of the other two classes and always errs in clas-sifying the test instance The leave-one-out estimatedaccuracy for a majont> inducer on the ins dataset istherefore 0% M oreover all folds have this estimated ac-curacy, thus the standard deviation of the folds is again0 %giving the unjustified assurance that 'he estimate is stable | The example shows an inherent problem with cross-validation th-t applies to more than just a majority in-ducer In a no-infornirition dataset where the label val-ues are completely random, the best an induction algo-rithm can do is predict majority Leave-one-out on such a dataset with 50% of the labels for each class and a majontv ind'-cer (the best, possible inducer) would still predict 0% accuracy 2 3 B o o t s t r a pThe bootstrap family was introduced by Efron and is fully described in Efron &. Tibshirani (1993) Given a dataset of size n a b o o t s t r a p s a m p l e is created by sampling n instances uniformly from the data (with re-placement) Since the dataset is sampled with replace-ment, the probability of any given instance not beingchosen after n samples is theKOHAVI 1139expected number of distinct instances from the original dataset appearing in the teat set is thus 0 632n The eO accuracy estimate is derived by using the bootstrap sam-ple for training and the rest of the instances for testing Given a number b, the number of bootstrap samples, let e0, be the accuracy estimate for bootstrap sample i The632 bootstrap estimate is defined as(5)where ace, is the resubstitution accuracy estimate on the full dataset (i e , the accuracy on the training set) The variance of the estimate can be determined by com puting the variance of the estimates for the samples The assumptions made by bootstrap are basically the same as that of cross-validation, i e , stability of the al-gorithm on the dataset the 'bootstrap world" should closely approximate the real world The b32 bootstrap fails (o give the expected result when the classifier is a perfect memonzer (e g an unpruned decision tree or a one nearest neighbor classifier) and the dataset is com-pletely random, say with two classes The resubstitution accuracy is 100%, and the eO accuracy is about 50% Plugging these into the bootstrap formula, one gets an estimated accuracy of about 68 4%, far from the real ac-curacy of 50% Bootstrap can be shown to fail if we add a memonzer module to any given inducer and adjust its predictions If the memonzer remembers the training set and makes the predictions when the test instance was a training instances, adjusting its predictions can make the resubstitution accuracy change from 0% to 100% and can thus bias the overall estimated accuracy in any direction we want3 Related W o r kSome experimental studies comparing different accuracy estimation methods have been previously done but most of them were on artificial or small datasets We now describe some of these effortsEfron (1983) conducted five sampling experiments and compared leave-one-out cross-validation, several variants of bootstrap, and several other methods The purpose of the experiments was to 'investigate some related es-timators, which seem to offer considerably improved es-timation in small samples ' The results indicate that leave-one-out cross-validation gives nearly unbiased esti-mates of the accuracy, but often with unacceptably high variability, particularly for small samples, and that the 632 bootstrap performed bestBreiman et al (1984) conducted experiments using cross-validation for decision tree pruning They chose ten-fold cross-validation for the CART program and claimed it was satisfactory for choosing the correct tree They claimed that "the difference in the cross-validation estimates of the risks of two rules tends to be much more accurate than the two estimates themselves "Jain, Dubes fa Chen (1987) compared the performance of the t0 bootstrap and leave-one-out cross-validation on nearest neighbor classifiers Using artificial data and claimed that the confidence interval of the bootstrap estimator is smaller than that of leave-one-out Weiss (1991) followed similar lines and compared stratified cross-validation and two bootstrap methods with near-est neighbor classifiers His results were that stratified two-fold cross validation is relatively low variance and superior to leave-one-outBreiman fa Spector (1992) conducted a feature sub-set selection experiments for regression, and compared leave-one-out cross-validation, A:-fold cross-validation for various k, stratified K-fold cross-validation, bias-corrected bootstrap, and partial cross-validation (not discussed here) Tests were done on artificial datasets with 60 and 160 instances The behavior observed was (1) the leave-one-out has low bias and RMS (root mean square) error whereas two-fold and five-fold cross-validation have larger bias and RMS error only at models with many features, (2) the pessimistic bias of ten-fold cross-validation at small samples was significantly re-duced for the samples of size 160 (3) for model selection, ten-fold cross-validation is better than leave-one-out Bailey fa E lkan (1993) compared leave-one-out cross-ahdation to 632 bootstrap using the FOIL inducer and four synthetic datasets involving Boolean concepts They observed high variability and little bias in the leave-one-out estimates, and low variability but large bias in the 632 estimatesWeiss and Indurkyha (Weiss fa Indurkhya 1994) con-ducted experiments on real world data Lo determine the applicability of cross-validation to decision tree pruning Their results were that for samples at least of size 200 using stratified ten-fold cross-validation to choose the amount of pruning yields unbiased trees (with respect to their optimal size) 4 M e t h o d o l o g yIn order to conduct a large-scale experiment we decided to use 04 5 and a Naive Bayesian classifier The C4 5 algorithm (Quinlan 1993) is a descendent of ID3 that builds decision trees top-down The Naive-Bayesian clas-sifier (Langley, Iba fa Thompson 1992) used was the one implemented in (Kohavi, John, Long, Manley fa Pfleger 1994) that uses the observed ratios for nominal features and assumes a Gaussian distribution for contin-uous features The exact details are not crucial for this paper because we are interested in the behavior of the accuracy estimation methods more than the internals of the induction algorithms The underlying hypothe-sis spaces—decision trees for C4 5 and summary statis-tics for Naive-Bayes—are different enough that we hope conclusions based on these two induction algorithms will apply to other induction algorithmsBecause the target concept is unknown for real-world1140 LEARNINGconcepts, we used the holdout method to estimate the quality of the cross-validation and bootstrap estimates To choose & set of datasets, we looked at the learning curves for C4 5 and Najve-Bayes for most of the super-vised classification dataaets at the UC Irvine repository (Murphy & Aha 1994) that contained more than 500 instances (about 25 such datasets) We felt that a min-imum of 500 instances were required for testing While the true accuracies of a real dataset cannot be computed because we do not know the target concept, we can esti mate the true accuracies using the holdout method The "true' accuracy estimates in Table 1 were computed by taking a random sample of the given size computing the accuracy using the rest of the dataset as a test set, and repeating 500 timesWe chose six datasets from a wide variety of domains, such that the learning curve for both algorithms did not flatten out too early that is, before one hundred instances We also added a no inform a tion d l stt, rand, with 20 Boolean features and a Boolean random label On one dataset vehicle, the generalization accu-racy of the Naive-Bayes algorithm deteriorated hy morethan 4% as more instances were g;iven A similar phenomenon was observed on the shuttle dataset Such a phenomenon was predicted by Srhaffer and Wolpert (Schaffer 1994, Wolpert 1994), but we were surprised that it was observed on two real world datasetsTo see how well an Accuracy estimation method per forms we sampled instances from the dataset (uniformly without replacement) and created a training set of the desired size We then ran the induction algorihm on the training set and tested the classifier on the rest of the instances L E I the dataset This was repeated 50 times at points where the lea rning curve wa s sloping up The same folds in cross-validation and the same samples m bootstrap were used for both algorithms compared5 Results and DiscussionWe now show the experimental results and discuss their significance We begin with a discussion of the bias in the estimation methods and follow with a discussion of the variance Due to lack of space, we omit some graphs for the Naive-Bayes algorithm when the behavior is ap-proximately the same as that of C 4 5 5 1 T h e B i a sThe bias of a method to estimate a parameter 0 is de-fined as the expected value minus the estimated value An unbiased estimation method is a method that has zero bias Figure 1 shows the bias and variance of k-fold cross-validation on several datasets (the breast cancer dataset is not shown)The diagrams clearly show that k-fold cross-validation is pessimistically biased, especially for two and five folds For the learning curves that have a large derivative at the measurement point the pessimism in k-fold cross-Figure ] C'4 5 The bias of cross-validation with varying folds A negative K folds stands for leave k-out E rror bars are 95% confidence intervals for (he mean The gray regions indicate 95 % confidence intervals for the true ac curaries Note the different ranges for the accuracy axis validation for small k s is apparent Most of the esti-mates are reasonably good at 10 folds and at 20 folds they art almost unbiasedStratified cross validation (not shown) had similar be-havior, except for lower pessimism The estimated accu-racy for soybe an at 2 fold was 7% higher and at five-fold, 1 1% higher for vehicle at 2-fold, the accuracy was 2 8% higher and at five-fold 1 9% higher Thus stratification seems to be a less biased estimation methodFigure 2 shows the bias and variance for the b32 boot-strap accuracy estimation method Although the 632 bootstrap is almost unbiased for chess hypothyroid, and mushroom for both inducers it is highly biased for soy-bean with C'A 5, vehicle with both inducers and rand with both inducers The bias with C4 5 and vehicle is 9 8%5 2 The VarianceWhile a given method may have low bias, its perfor-mance (accuracy estimation in our case) may be poor due to high variance In the experiments above, we have formed confidence intervals by using the standard de-viation of the mea n a ccura cy We now switch to the standard deviation of the population i e , the expected standard deviation of a single accuracy estimation run In practice, if one dots a single cross-validation run the expected accuracy will be the mean reported above, but the standard deviation will be higher by a factor of V50, the number of runs we averaged in the experimentsKOHAVI 1141Table 1 True accuracy estimates for the datasets using C4 5 and Naive-Bayes classifiers at the chosen sample sizesFigure 2 C4 5 The bias of bootstrap with varying sam-ples Estimates are good for mushroom hypothyroid, and chess, but are extremely biased (optimistically) for vehicle and rand, and somewhat biased for soybeanIn what follows, all figures for standard deviation will be drawn with the same range for the standard devi-ation 0 to 7 5% Figure 3 shows the standard devia-tions for C4 5 and Naive Bayes using varying number of folds for cross-validation The results for stratified cross-validation were similar with slightly lower variance Figure 4 shows the same information for 632 bootstrap Cross-validation has high variance at 2-folds on both C4 5 and Naive-Bayes On C4 5, there is high variance at the high-ends too—at leave-one-out and leave-two-out—for three files out of the seven datasets Stratifica-tion reduces the variance slightly, and thus seems to be uniformly better than cross-validation, both for bias and vananceFigure 3 Cross-validation standard deviation of accu-racy (population) Different, line styles are used to help differentiate between curves6 S u m m a r yWe reviewed common accuracy estimation methods in-cluding holdout, cross-validation, and bootstrap, and showed examples where each one fails to produce a good estimate We have compared the latter two approaches on a variety of real-world datasets with differing charac-teristicsProposition 1 shows that if the induction algorithm is stable for a given dataset, the variance of the cross-validation estimates should be approximately the same, independent of the number of folds Although the induc-tion algorithms are not stable, they are approximately stable it fold cross-validation with moderate k values (10-20) reduces the variance while increasing the bias As k decreases (2-5) and the sample sizes get smaller, there is variance due to the instability of the training1142 LEARNING1 igure 4 632 Bootstrap standard deviation in acc-rat y (population)sets themselves leading to an increase in variance this is most apparent for datasets with many categories, such as soybean In these situations) stratification seems to help, but -epeated runs may be a better approach Our results indicate that stratification is generally a better scheme both in terms of bias and variance whencompared to regular cross-validation Bootstrap has low,variance but extremely large bias on some problems We recommend using stratified Len fold cross-validation for model selection A c k n o w l e d g m e n t s We thank David Wolpert for a thorough reading of this paper and many interesting dis-cussions We thank Tom Bylander Brad E fron Jerry Friedman, Rob Holte George John Pat Langley Hob Tibshiram and Sholom Weiss for their helpful com nients and suggestions Dan Sommcrfield implemented Lhe bootstrap method in WLC++ All experiments were conducted using M L C ++ partly partly funded by ONR grant N00014-94-1-0448 and NSF grants IRI 9116399 and IRI-941306ReferencesBailey, T L & E lkan C (1993) stimating the atcuracy of learned concepts, in Proceedings of In ternational Joint Conference on Artificial Intelli-gence , Morgan Kaufmann Publishers, pp 895 900 Breiman, L & Spector, P (1992) Submodel selectionand evaluation in regression the x random case Inttrnational St atistic al Review 60(3), 291-319 Breiman, L , Friedman, J H , Olshen, R A & StoneC J (1984), Cl a ssific ation a nd Regression Trets Wadsworth International GroupEfron, B (1983), 'E stimating the error rate of a pre-diction rule improvement on cross-validation",Journal of the Americ an St atistic al Associ ation 78(382), 316-330 Efron, B & Tibshiram, R (1993) An introduction tothe bootstra p, Chapman & HallJam, A K Dubes R C & Chen, C (1987), "Boot-strap techniques lor error estimation", IEEE tra ns-actions on p a ttern a n a lysis a nd m a chine intelli-gence P A M I -9(5), 628-633 Kohavi, R , John, G , Long, R , Manley, D &Pfleger K (1994), M L C ++ A machine learn-ing library in C ++ in 'Tools with Artifi-cial Intelligence I E E EComputer Society Press, pp 740-743 Available by anonymous ftp from s t a r r y Stanford E DU pub/ronnyk/mlc/ toolsmlc psLangley, P Tba, W & Thompson, K (1992), An anal-ysis of bayesian classifiers in Proceedings of the tenth national conference on artificial intelligence",A A A I Press and M I T Press, pp 223-228Murph' P M & Aha D W (1994), V( I repository of machine learning databases, For information con-tact ml-repository (Ui(,s uci edu Quinlan I R (1993) C4 5 Progra ms for Ma chine Learning Morgan Kaufmann Los Altos CaliforniaSchaffcr C (19941 A conservation law for generalization performance, in Maehinc Learning Proceedings of Lhe E leventh International conference Morgan Kaufmann, pp 259-265Shao, J (1993), Linear model seletion via cross-validation Journ a l of the America n sta tistica l As-sociation 88(422) 486-494 Weiss S M (1991), Small sample error rate estimationfor k nearest neighbor classifiers' I E EE Tr a ns ac tions on Pa ttern An alysis a nd Ma chine Intelligence 13(3), 285-289 Weiss, S M & lndurkhya N (1994) Decision Lreepruning Biased or optimal, in Proceedings of the twelfth national conference on artificial intel-ligence A A A I Press and M I T Press pp 626-632 Wolpert D H (1992), Stacked generalization , Neura lNetworks 5 241-259 Wolpert D H (1994a) Off training set error and a pri-ori distinctions between learning algorithms, tech-mcal Report SFI TR 94-12-121, The Sante Fe ln-stituteWolpert D II {1994b), The relationship between PAC, the statistical physics framework the Bayesian framework, and the VC framework Technical re-port, The Santa Fe Institute Santa Fe, NMZhang, P (1992), 'On the distributional properties of model selection criteria' Journ al of the America nStatistical Associa tion 87(419), 732-737 KOHAVI 1143。

简介ACL(Access Control List 访问控制列表)是路由器和交换机接口的指令列表，用来控制端口进出的数据包。

ACL的定义也是基于每一种被动路由协议的，且适用于所有的被动路由协议(如IP、IPX、Apple Talk等)，如果路由器接口配置成为三种协议(IP、Apple Talk和IPX)，那么必须定义三种ACL来分别控制这三种协议的数据包。

ACL的作用ACL可以限制网络流量、提高网络性能；提供网络安全访问的基本手段；可以在路由器端口处决定哪种类型的通信流量被转发或被阻塞，应用范围很广，如：路由过滤、Qos、NAT、Router-map、VTY等。

ACL的分类根据过滤层次：基于IP的ACL(IP ACL)、基于MAC的ACL(MAC ACL)，专家ACL(Expert ACL)。

根据过滤字段：标准ACL(IP ACL、MAC ACL)、扩展ACL(IP ACL、MAC ACL、专家ACL)。

根据命名规则：表号ACL、命名ACL。

说明：标准型ACL功能非常简单；而扩展型ACL功能非常强大，匹配的更详细，对路由器等网络设备的性能要求更高，或者对于网速的拖慢更明显，组网时需要酌情使用。

ACL的工作原理ACL的执行按照列表中条件语句的顺序从上到下、逐条依次判断执行。

如果一个数据包的报头跟表中某个条件判断语句相匹配，则不论是第一条还是最后一条语句，数据包都会立即发送到目的端口，那么后面的语句将被忽略，不再进行检查。

当数据包与前一个判断条件不匹配时，才被交给下一条判断语句进行比较。

如果所有的ACL判断语句都检测完毕，ACL遵循的规范和原则1.ACL的列表号指出了是哪种协议的ACL。

各种协议有自己的ACL，而每个协议的ACL 又分为标准ACL和扩展ACL。

这些ACL都是通过ACL表号区别的。

如果在使用一种访问ACL时用错了列表号，那么就会出错。

2.一个ACL的配置是基于每种协议的每个接口的每个方向。