Proceedings of the 1st Iberian Conference on Pattern Recognition and Image Analysis, IbPRIA 2003? Lecture Notes in Computer Science Vol. 2652. Springer 2003
Learning decision trees and tree automata for a
syntactic pattern recognition task
Jos′e M.Sempere and Dami′a n L′o pez
Departamento de Sistemas Inform′a ticos y Computaci′o n
Universidad Polit′e cnica de Valencia,Valencia(Spain)
email:{jsempere,dlopez}@dsic.upv.es
Abstract.Decision trees have been widely used for di?erent tasks in
arti?cial intelligence and data mining.Tree automata have been used
in pattern recognition tasks to represent some features of objects to be
classi?ed.Here we propose a method that combines both approaches to
solve a classical problem in pattern recognition such as Optical Charac-
ter Recognition.We propose a method which is organized in two stages:
(1)we use a grammatical inference technique to represent some struc-
tural features of the characters and,(2)we obtain edit distances between
characters in order to design a decision tree.The combination of both
methods bene?ts from their individual characteristics and is formulated
as a coherent unifying strategy.
1Introduction
Syntactic Pattern Recognition is a well known research area from Arti?-
cial Intelligence in which the target task is to recognize objects from the
real world(speech,image,medical signals,...)which are represented as
formal languages[HU79].Mainly,the most common representations in
these tasks have been some families of string languages(regular,context-
free,...),some families of tree languages(regular ones)or some families
of graph languages(graphs based on vertex substitutions or hypergraphs
with edge replacement).So,the goal in any syntactic pattern recogni-
tion learning task is to guess the hidden formal language from examples
(strings,trees or graphs).
By the other hand,decision trees[Qu93]can be considered as tree-like
representations of?nite sets of if-then-else rules.This representation a-
llows to take some decisions for the analysis of a set of attributes of a
given concept.Mainly,the decision can be applied to a classi?cation task,
a predictive task or an advisement task(i.e.expert systems).During the
last years,decision trees have been applied in the very promising area of
data mining[MBK98]to extract knowledge from large databases.
In this work,we combine these two di?erent approaches to the learning
problem in order to construct a system to solve an Optical Character
Recognition(OCR)task.Here,we will work only with handwritten iso-
lated digits from0to9.Our solution is based on a two stages system.
Work supported by the Spanish CICYT under contract TIC2000-1153.
First,the system learns a set of tree automata(one per digit)by using an error-correcting technique based on a grammatical inference method. Basic concepts and methods on grammatical inference can be viewed in [AS83,Sa97].Then,the system obtains a set of edit distances of every digit to every tree automaton.In the last stage,the system learns a decision tree from the last set of distances that will classify any digit according to a set of rules based on distances.
The structure of this work is as follows:First,we will explain the OCR task that the method attempts to solve.We will explain the learning methods on every phase(i.e.learning of tree automata and learning of decision trees).Finally,we will show some preliminary results from an experimentation using our approach and we will give some research guidelines for future works.
2The problem:Optical Character Recognition The target problem of this work is related to the working area of Hand-written Recognition.Here,the general goal is to construct a robust sys-tem which be able to recognize any phrase or text that has been pre-viously handwritten by a human being.This task has not yet completely solved.So,some subproblems are involved to solve this task.For ex-ample,there exists an increasing area that attempts to construct good segmentation rules in order to factorize any phrase in words an any word in letters or digits.Other researchers have focused their interest on cons-tructing good language models for task-oriented systems(for example, some systems are focused on medical writings,or mathematics writings and so on).We will focus on another task which consists on isolated dig-its recognition.The solution to this task is important to construct more sophisticated systems.Here,the task is quite simple given that phrase and word segmentation tasks are avoided.This is the problem that we try to solve with a syntactic pattern recognition approach.
2.1Representation of the digits
First,we will consider how the real world objects will be represented. Let us observe in Figure1a digit2that has been obtained from a handwriting scanning.
Under our approach,the?rst stage to represent any digit is to obtain a quad tree(qtree)[HS79]from its digital image.A qtree can be cons-tructed by drawing a square window around the digit and splitting the window in four windows of the same size recursively up to a prede?ned depth.In Figure2we can observe how the window of digit2is recursively split.
Once the system obtains the windows of the digit,then it assigns a label to every window of the smallest size.The systems assigns one label to e-very window depending on the grey scale(black,white or grey).So,every smallest window is represented by a label of a three symbols alphabet (i.e.{a,b,c}).The relationships between windows can be represented by a tree by using an up-down and left-to-right scanning of the qtree.In
Fig.1.Handwritten digit2.
Fig.2.Constructing a qtree for digit2.
Figure3we can observe the tree obtained for digit2by using a depth that equals to3while constructing the qtree.From now on,we will use this tree representation.
3Learning methods
We will use two di?erent learning paradigms to solve the learning stage of the previously de?ned problem.First,we will use grammatical inference methods to construct a tree automaton from every set of trees represen-ting the same digit.Then,we will go to a second learning stage to obtain a di?erent representation of the digit based on distances of every digit (every tree)to every concept(every automaton).From this second repre-sentation we will infer a decision tree by using standard methods based on the entropy of the examples and distances.The learning scheme is showed in Figure4.
Now we will explain the di?erent methods that we have used at every learning stage.
3.1Grammatical Inference of error-correcting tree automata
The?rst stage of our learning approach is based on a grammatical infer-ence method for tree languages.Grammatical inference[AS83,Sa97]is an
σ
b b b b b b b b b b b b b b b b b b b b b b
Fig.3.Handwritten digit qtree with a depth that equals https://www.doczj.com/doc/7e11236143.html,bel a corresponds to a at least75%white square,label b corresponds to a at least75%black square,and label c corresponds to a grey square.
inductive approach to the learning problem based on the representation
of concepts as formal languages.Here,as previously explained,we use
trees to represent the digits for the OCR task.
Several methods have been proposed to infer tree languages from exam-
ples[Ga93,GO93,Sa92].We will apply a method based on error-correcting
distances from trees to tree automata.The de?nition of such distance is
based on classical editing distances for strings to?nite string automata
[LSG00].Once,the distance has been de?ned then,the learning method
is an error-correcting grammatical inference technique[LE02].
3.2C
4.5learning algorithm
Learning decision trees is a classical topic on machine learning.A decision
tree is a representation of a?nite set of if-then-else rules.The main
characteristics of decision trees are the following:
1.The examples can be de?ned as a set of numerical and symbolic
attributes.
2.The examples can be incomplete or contain noisy data.
3.The main learning algorithms work under Occam’s razor principle
and Minimum Description Length approaches.
The main learning algorithms for decision trees have been proposed by
Quinlan[Qu93].First,Quinlan de?ned ID3algorithm based on the in-
formation gain principle.This criterion is performed by calculating the
entropy that produces every attribute of the examples and by selec-
ting the attributes that save more decisions in information https://www.doczj.com/doc/7e11236143.html,ter,
Quinlan de?ned C4.5algorithm[Qu93]which is an evolution of ID3al-
gorithm.We will use C4.5algorithm for the second learning phase.The
main characteristics of C4.5are the following:
1.The algorithm can works with continuous attributes(i.e.real data).
https://www.doczj.com/doc/7e11236143.html,rmation gain is not the only learning criterion.
3.The trees can be post-pruned in order to re?ne the desired output.
4Experiments and results
We have performed two experiments in order to carry out a?rst eva-
luation of our learning strategy.The digits that we have used for training
Fig.4.Our learning strategy to solve the OCR problem.
and test is a subset from the data set”NIST SPECIAL DATABASE3, NIST Binary Images of Handwritten Segmented Characters”[Ga94].
The protocol that we have performed in both experiments is the following one:First,we obtain the qtree representations of every digit in the data set.Then,we divide this set in two disjoint subsets(Set1and Set2) and we apply to Set1the Error-Correcting inference technique in order to obtain a tree automaton for every digit.We calculate the distance of every digit to every automaton(so,every digit has ten attributes that represent the distances to every model).Then,we calculate the distances of every digit in Set2to every tree automaton.
From Set1and Set2we perform a learning plus testing phase for decision trees.We have used an implementation of C4.5algorithm in C which is available from internet at J.R.Quinlan’s Home Page[QuHTTP].Observe that for every digit at Set1there is at least one distance with value0, while this is not true in general for digits of Set2.
Experiment1
We have selected3000digits for Set1(300samples for every digit)and 1000digits for Set2(100samples for every digit).We have performed three rounds on C4.5algorithm in order to use di?erent samples with or without distance0.The results of this experiment are showed in Figure 5.
Round1Round2 Evaluation on training data(2666items)
Before Pruning After Pruning
Size Errors 18944(1.7%)Size Errors Estimate 17348(1.8%) 5.8%
Evaluation on test data(1334items) Before Pruning After Pruning
Size Errors 189122(9.1%)Size Errors Estimate
173120(9.0%) 5.8%
Evaluation on training data(2667items)
Before Pruning After Pruning
Size Errors
16754(2.0%)
Size Errors Estimate
15955(2.1%)(5.7%)
Evaluation on test data(1333items)
Before Pruning After Pruning
Size Errors
167116(8.7%)
Size Errors Estimate
159114(8.6%)(5.7%)
Round3
Evaluation on training data(2667items)
Before Pruning After Pruning
Size Errors
20556(2.1%)
Size Errors Estimate
18760(2.2%) 6.5%
Evaluation on test data(1333items)
Before Pruning After Pruning
Size Errors
20587(6.5%)
Size Errors Estimate
18786(6.5%) 6.5%
Fig.5.Results for the?rst experiment.
Experiment2
We have selected3000digits for Set1(300samples for every digit)and 2000digits for Set2(200samples for every digit).We have performed three rounds on C4.5algorithm in order to use di?erent samples with or without distance0.The results of this experiment are showed in Figure 6.
Round1Round2 Evaluation on training data(3333items)
Before Pruning After Pruning
Size Errors 31588(2.6%)Size Errors Estimate 29393(2.8%)8.1%
Evaluation on test data(1667items) Before Pruning After Pruning
Size Errors 315189(11.3%)Size Errors Estimate
293185(11.1%)8.1%
Evaluation on training data(3333items)
Before Pruning After Pruning
Size Errors
30586(2.6%)
Size Errors Estimate
28193(2.8%)7.9%
Evaluation on test data(1667items)
Before Pruning After Pruning
Size Errors
305186(11.2%)
Size Errors Estimate
281185(11.1%)7.9%
Round3
Evaluation on training data(3334items)
Before Pruning After Pruning
Size Errors
32388(2.6%)
Size Errors Estimate
28997(2.9%)8.1%
Evaluation on test data(1666items)
Before Pruning After Pruning
Size Errors
323207(12.4%)
Size Errors Estimate
289208(12.5%)8.1%
Fig.6.Results for the second experiment.
Conclusions
It can be observed that,for every round that we have carried out on the experiments,the error median in training data is less than the one in test data.This is a trivial result that all learning methods would hold. After,the pruning of the decision trees the median error decreases.It implies that some rules that C4.5obtains are not useful for the classi?-cation task.
The results on the?rst experiment are better than in the second.Here, the input sample de?nes how the rules are extracted.In the?rst expe-riment,there is a number of examples with distance0to any tree au-tomata which is three times those examples whose distances to every tree automata is not equal to0.So,the input sample for constructing the tree automata is very important to obtain not only the distances but the decision tree.
Finally,an important remark is that the method has a better perfor-mance than some other methods that uses only a grammatical inference approach[LE02].Furthermore,if we compare this method with some other methods based on geometrical approaches then,the di?erences between median errors can be balanced with the complexity behaviors (i.e.geometrical methods have a worst behavior than our approach under time and space complexities).
5Future works
From the initial results that we have obtained,our approach to the OCR problem has showed itself as a promising one.Anyway,we can point out to the following research guidelines in order to improve this work.
–We should enrich the attributes of every digit by including not only the distances but some other structural features.
–The criteria for decision tree learning could be change in order to take into account the distribution of the distances obtained from tree automata.
–Finally,we should apply this method to other pattern recognition tasks.
References
[AS83] D.Angluin,C.Smith.Inductive Inference:Theory and Methods.
Computing Surveys,vol.15.No.3,pp237-269.1983.
[Ga93]P.Garc′?a.Learning k-testable tree sets from positive data.Tech-nical Report DSIC II/46/1993.Departamento de Sistemas In-form′a ticos y Computaci′o n.Universidad Polit′e cnica de Valencia.
1993.
[GO93]P.Garc′?a and J.Oncina.Inference of recognizable tree sets.
Technical Report DSIC II/47/1993.Departamento de Sistemas In-form′a ticos y Computaci′o n.Universidad Polit′e cnica de Valencia.
1993.
[Ga94]M.D.Garris.Design and Collection of a handwriting sample im-age database.Encycl.of Comp.Sci.&Tech.Marcel Dekker,N.Y.
Vol31,Supp.16,pp189-214.1994.
[HU79]J.Hopcroft,J.Ullman.Introduction to Automata Theory,Lan-guages and Computation.Addison-Wesley Publishing Co.1979. [HS79]G.M.Hunter and K.Steiglitz.Operations on images using quad trees.IEEE Transactions on Pattern Analysis and Machine Intelli-gence Vol.1,No.2,pp145-153.1979.
[LSG00] D.L′o pez,J.M.Sempere,P.Garc′?a.Error Correcting Analysis for Tree Languages.International Journal on Pattern Recognition and Arti?cial Intelligence,Vol.14,No.3,pp357-368.2000.
[LE02] D.L′o pez,S.Espa?n a.Error-correcting tree language inference.
Pattern Recognition Letters23,pp1-12.2002.
[MBK98]R.Michalski,I.Bratko and M.Kubat.Machine Learning and Data Mining.Methods and Applications.John Wiley and Sons LTD.
1998.
[Qu93]J.R.Quinlan.C4.5:programs for machine learning.Morgan Kaufmann.1993.
[QuHTTP]R.Quinlan’s Home Page
https://www.doczj.com/doc/7e11236143.html,.au/～quinlan/
[Sa92]Y.Sakakibara.E?cient learning of context-free grammars from positive structural https://www.doczj.com/doc/7e11236143.html,rmation and Computation97,pp 23-60.1992.
[Sa97]Y.Sakakibara.Recent advances of grammatical inference.Theo-retical Computer Science185,pp15-45.1997.
生化常见异常结果及原因分析 反应曲线波动、跳动 一、几乎所有反应曲线跳动 1、电源接地不良：主电源接地不良、光电盒接地不良； 2、反应盘进水：真空泵压力密封圈磨损、真空泵电源接触不良、管接头处漏气、管路漏气、吸废液钢管堵塞、单向阀坏、比色杯破裂； 3、搅拌杆不搅拌：搅拌电机坏、搅拌电机线接触不良、搅拌杆顶住比色杯底； 4、光路歪：灯泡装歪、比色杯装歪、温控锅装歪、透镜装歪； 5、试剂加入异常：试剂针堵塞、试剂注射器脱落、试剂针接头处脱落、快速接头脱落； 6、一次性比色杯不干净；
7、阳光直射反应光电系统； 8、外界干扰因素：如发动机，电钻等。 二、只有个别波长反应曲线跳动 电源接地不良：主电源接地不良、光电盒接地不良。 三、只有个别项目反应曲线跳动 试剂异常：试剂变混浊、变色或试剂放错位置； 试剂经搅拌后起较多气泡，挡住光路。 反应曲线异常 一、几乎所有反应曲线形状正确，反应曲线平稳，但基本无反应 样本未加入：样本针堵塞、样本注射器脱落、样本针接头处脱落、快速接头脱落。
二、个别项目反应曲线形状正确，反应曲线平稳，但基本无反应 第二试剂未加入：第二试剂放错位置 三、个别项目反应曲线形状改变，但反应曲线平稳 试剂加入异常：试剂盘固定销钉脱落、试剂注射器漏气。 四、个别项目反应曲线形状改变，剧烈上升或下降，反应曲线或平稳，或小幅波动 试剂异常：试剂性能差或失效，最易发生在ALP、GGT、AMY等项目上。 五、个别测试反应曲线形状改变，剧烈上升或下降，但反应曲线平稳 1、试剂间交叉污染：最易发生在TG、TC、Glu和Bun等项目上；
2、样本异常：样本中含有某些药物或干扰成分，最易发生在TB、DB等项目上。 反应曲线正常，结果重复性差 一、几乎所有项目重复性差 1、样本加入异常：样本针半堵塞、样本注射器中有气泡、样本针接头处密封不严、样本针内壁阀关闭不严； 2、试剂加入异常：试剂针半堵塞、试剂注射器中有气泡、试剂针接头处密封不严、试剂针内壁阀关闭不严； 3、搅拌异常：搅拌电机坏、搅拌电机线接触不良、搅拌杆顶住比色杯底； 4、试剂/样本内外壁清洗不够，交叉污染大：液泵压力下降、第二级过滤器堵塞； 5、反应盘温度不稳定：反应盘温度波动较大。 二、个别项目重复性差
感受现代科技 【学习目标】 1、知识：感受现代科技给人类生活带来的新变化，认识科技与生活，科技发展与社会发展的关系，懂得“科学技术是第一生产力”的道理。 2、能力与情感：感悟现代科技的神奇与力量，理解科技是社会发展的强大推力，激发学生 对科技重要性的认识，增强学生对科学的兴趣，培养学生热爱科学的精神。 3、过程与方法：依据教学内容和学生的认识规律设置了“课前预习”、“课堂助学”、“课堂巩固”、“课后拓学”、“教学反思”五个模块的教学整合，运用多媒体等教学手段，采用自主体验、 探究活动、案例情境等方法来完成教学目标。 【学习重点、难点】 领略现代科技的神奇与力量，理解“科技是第一生产力”。 【学习过程】 一、预习初探： （一）快快行动，书外的知识真有趣： 1、生活体验：观察生活，请你说说我们身边有哪些科技产品？例举实例说说这些科技产品给我们的生活带来哪些新变化？ 2、想象天地：展现你的想象天份，想象你准备发明一样科技产品，使你的未来生活更美好。 3、图片收集：上网收集有关科技产品的图片，准备创办科技小展览，领略现代科技的神 奇与力量。 （二）阅读课本，书本的知识真寻味： 4、我们现在的生活与科技________________。丰足的衣食，舒适的住行，千百年来一直是人类_________________。 5、科学技术是________________的强大推力，是________生产力。______________已成为当代经济发展的火车头。 6、________________是人类文明的标志。科学技术的进步为人类创造了巨大的 ______________和_________________。
临床生化检验中反应曲线的应用分析 发表时间：2018-07-03T17:25:04.580Z 来源：《中国研究型医院》2018年5卷2期作者：栾春红 [导读] 临床生化检验中应用反应曲线，不仅可以提高生化检验结果的准确性，而且可以节省检验时间。 黑龙江省伊春市桃山林业局职工医院黑龙江伊春市 152514 【摘要】目的探讨反应曲线在临床生化检验中的应用效果。方法选择2016年7月至2017年7月来我院进行生化检验的患者122例，随机分为观察组和对照组两组，每组61例，对照组采用常规生化检验和检验报告，观察组在与对照组相同的基础上采用反应曲线，比较两组检验结果准确率及检验时间。结果观察组检验准确率为93.44%，显著高于对照组的54.10%，差异有统计学意义（P<0.05）；观察组检验时间显著少于对照组，差异有统计学意义（P<0.05）。结论临床生化检验中应用反应曲线，不仅可以提高生化检验结果的准确性，而且可以节省检验时间。 【关键词】反应曲线；临床生化检验；应用效果 Application Analysis of Response Curve in Clinical Biochemical Examination Objective To investigate the effect of reaction curve in clinical biochemical examination. Methods 122 patients who came to our hospital from July 2016 to July 2017 for biochemical examination were randomly divided into two groups, the observation group and the control group, each group with 61 cases. The control group used the conventional biochemical test and test report. The observation group used the reaction curve on the same basis as the control group. The accuracy and time of the test results were compared between the two groups. Results The accuracy of the test in the observation group was 93.44 %, which was significantly higher than the control group's 54.10 %. The difference was statistically significant(P<0.05） In the observation group, the test time was significantly less than that of the control group, and the difference was statistically significant(P<0.05） Conclusion The application of response curve in clinical biochemical test can not only improve the accuracy of biochemical test results, but also save the test time. [Keywords] reaction curve; Clinical biochemical examination; Application Effect 全自动生化分析仪是临床上应用十分广泛的生化检验仪器，该仪器在临床检验中的应用不仅有效提高了临床检验工作效率，同时也促使临床检验工作更加趋于标准化和规范化，在很大程度上促进了临床生化检验的发展[1]。但在临床实际生化检验工作中，存在多数检验人员只对其进行常规检测和维护保养，而忽略了对其实时曲线检测系统的应用。笔者通过对该院生化检验者采用反应曲线，并与同期采用常规生化检验和检验报告者进行了比较，现做如下报告： 1资料与方法 1.1一般资料选择2016年7月至2017年7月来我院进行生化检验的患者122例，其中男性66例，女性56例，年龄14-54岁，平均年龄（43.9±6.5）岁；随机分为观察组和对照组两组，每组61例，两组患者在年龄、性别等一般资料方面比较，差异无统计学意义，P>0.05，具有可比性。 1.2方法运用贝克曼AU5800型全自动生化分析仪和北京利德曼公司生产的检测试剂和校准品，定期开展质控品检测，以提高检测结果的控制力。对照组采用常规生化检验和检验报告，由临床医务人员来分析生化检验报告，对疾病进行判断和确定。观察组在与对照组相同的基础上采用反应曲线，临床医务人员应认真观察生化检验中空白试剂、质控品和标准品的反应曲线，分析生化检验反应曲线出现异常的原因，让检验结果得以有效提升。 1.3观察指标比较两组检验准确率及检验时间。 1.4统计学方法采用SPSS17.0软件对数据进行统计分析，对计数资料率的比较采用x2检验，计量资料组间比较采用t检验，当P<0.05时，为差异有统计学意义。 2结果 2.1两组检验准确率比较如表1所示，观察组检验准确率为9 3.44%，显著高于对照组的5 4.10%，差异有统计学意义（P<0.05）。 表1 两组检验准确率比较（例，%） 2.2两组检验时间比较如表2所示，观察组检验时间显著少于对照组，差异有统计学意义（P<0.05）。 表2 两组检验时间比较 3讨论 在临床生化检验中应用全自动生化分析仪能让检验工作效率显著提高，但是由于机械设备机械化工作的影响，生化检验的准确率较差，通过反应曲线能对生化检验中的异常情况进行及时发现，进而来制定科学、合理的改善对策。在检验碱性磷酸酶时，检验结果为 224U/L，分析反应曲线发现，在加入试剂后反应曲线表现为一定程度的下降， 因此需重新开展碱性磷酸酶检验，第二次的碱性磷酸酶检验结果为77U/L，反应曲线无异常，结果显示在首次生化检验中仪器设备的状态不稳定；在胆红素检验中，反应曲线表现为波浪状，和该时段其他检测项目的检验结果类似，出现这种情况可能是因为电源灯在该时段内的使用时间较长，出现电源灯不稳定或者老化现象，所以应对电源灯进行及时更换。在进行碱性磷酸酶检验时，试剂空白处的吸光度高达21241，但是试剂说明书规定生化检验中所用试剂的空白处吸光度不能大于8000，结果显示试剂发生了变质。因此在临床生化检验中，应连续监测试剂空白处的吸光度，同时结合该季度反应曲线，进而来对曲线稳定性进行了解，让生化检验的结果准确性得以保证[2]。部分标
第十四课感受现代科技 项目一现代科技在身边 教师寄语：今日沟通于昔日之最大差异：在于科技的介入，已超越时间、空间，甚至权利与阶级的围墙。 学习目标： ●情感态度价值观：感受现代科技发展带来的新变化，培养学生热爱科学的情感和品质。 ●能力：培养学生观察、收集、整理、归纳信息的能力。 ●知识：了解科技对社会发展具有推动作用，理解科学技术是第一生产力。 重难点：科技是社会发展的强大推动力 学习过程： 一、课前预习（提前预习课本，顺便做以下小题，相信你能完成!） 1.我们现在的生活与科技_________。 2.___________________是第一生产力。 3._______________已成为当代经济发展的火车头。 4.科学技术是_______________的标志。 5.科技是社会发展的______________。 二、合作探究，共同进步（合作有助于提高学习效率，要努力哦！） 知识点一：现代科技给生活带来新变化 1：（见教材P58页-59页）图片1：液晶彩色电视图片2：利用网络学习 图片3：山里人用上了手机图片4：磁悬浮列车 说一说：观察生活，列举实例，说说现代科技给我们的生活带来哪些新的变化？ 2：阅读教材（P59页教材正文），说说现代科技的发展给人类的生活产生了怎样的影响？（学生讨论交流） 。3：（见教材P59页-60页）想象和推测一下，随着科技的发展，5年、10年、20年、50年后人们的生活可能是怎样的？ 。 知识点二：科技——社会发展的强大推力 1：（见教材P61页），材料中的数字变化说明了什么？（学生讨论交流） 上述材料中的数字变化说明：“”。 2：为什么说“科学技术是第一生产力”？ ① ② 实践证明：高新技术及其产业已经成为当代经济发展的火车头。 3：（见教材P62页）说一说：网络学校的发展将会给人们的学习方式带来了哪些变化？ 4、①科学技术的进步使精神文明建设有了新的载体. 说一说：在思想文化传播的载体方面，你知道有哪些新的传播手段呢？ ②科学技术的进步丰富了人们的 . 5、科学技术的进步为人类创造了巨大的财富和财富 三、课堂小结 通过学习我学会：
《交通规划》课程教学大纲 课程编号：E13D3330 课程中文名称：交通规划 课程英文名称：Transportation Planning 开课学期：秋季 学分/学时：2学分/32学时 先修课程：管理运筹学，概率与数理统计，交通工程学 建议后续课程：城市规划，交通管理与控制 适用专业/开课对象：交通运输类专业/3年级本科生 团队负责人：唐铁桥责任教授：执笔人：唐铁桥核准院长： 一、课程的性质、目的和任务 本课程授课对象为交通工程专业本科生，是该专业学生的必修专业课。通过本课程的学习，应该掌握交通规划的基础知识、常用方法与模型。课程具体内容包括：交通规划问题分析的一般方法，建模理论，交通规划过程与发展历史，交通调查、出行产生、分布、方式划分与交通分配的理论与技术实践，交通网络平衡与网络设计理论等，从而在交通规划与政策方面掌握宽广的知识和实际的操作技能。 本课程是一间理论和实践意义均很强的课程，课堂讲授要尽量做到理论联系实际，模型及其求解尽量结合实例，深入浅出，使学生掌握将交通规划模型应用于实际的基本方法。此外，考虑到西方在该领域内的研究水平，讲授时要多参考国外相关研究成果，多介绍专业术语的英文表达方法以及相关外文刊物。课程主要培养学生交通规划的基本知识、能力和技能。 二、课程内容、基本要求及学时分配 各章内容、要点、学时分配。适当详细，每章有一段描述。 第一章绪论（2学时） 1. 交通规划的基本概念、分类、内容、过程、发展历史、及研究展望。 2. 交通规划的基本概念、重要性、内容、过程、发展历史以及交通规划中存在的问题等。
第二章交通调查与数据分析（4学时） 1. 交通调查的概要、目的、作用和内容等；流量、密度和速度调查；交通延误和OD调查；交通调查抽样；交通调查新技术。 2. 交通中的基本概念，交通流量、速度和密度的调查方法，调查问卷设计与实施，调查抽样，调查结果的统计处理等。 第三章交通需求预测（4学时） 1. 交通发生与吸引的概念；出行率调查；发生与吸引交通量的预测；生成交通量预测、发生与吸引交通量预测。 2. 掌握交通分布的概念；分布交通量预测；分布交通量的概念，增长系数法及其算法。 3. 交通方式划分的概念；交通方式划分过程；交通方式划分模型。 第四章道路交通网络分析（4学时） 1. 交通网络计算机表示方法、邻接矩阵等 2. 交通阻抗函数、交叉口延误等。 第五章城市综合交通规划（2学时） 1. 综合交通规划的任务、内容；城市发展战略规划的基本内容和步骤 2. 城市中长期交通体系规划的内容、目标以及城市近期治理规划的目标与内容 第六章城市道路网规划（2学时） 城市路网、交叉口、横断面规划及评价方法。 第七章城市公共交通规划（2学时） 城市公共交通规划目标任务、规划方法、原则及技术指标。 第八章停车设施规划（2学时） 停车差设施规划目标、流程、方法和原则。 第九章城市交通管理规划（2学时） 城市交通管理规划目标、管理模式和管理策略。 第十章公路网规划（2学时） 公路网交通调查与需求预测、方案设计与优化。 第十一章交通规划的综合评价方法（2学时） 1. 交通综合评价的地位、作用及评价流程和指标。 2. 几种常见的评价方法。 第十二章案例教学（2学时）
走进都市农业感受现代科技 --"大东农业观光园"基地实践活动 设计者：大兴区第九小学窦雪征 指导教师：大兴区教师进修学校柏东河 ( 2008-05-27 13:40:19 ) 一、环境分析 我国是个农业大国。当前，在农业科技领域，中国不断缩小与发达国家的差距，农业科技部门在生物技术、高新技术、基础研究方面均取得较大进展，植物细胞和组织培养、花药培养、单倍体育种及其应用研究处于国际先进地位。但是，作为国家希望的小学生却对我国农业的了解知之甚少，甚至有的学生只认识餐桌上的蔬菜，田地里面的蔬菜就不认识了。而且由于社会环境的影响，学生对于农业生产的认识非常极端，认为农业生产只是辛苦的体力劳动，对其间蕴含的高科技文化知识基本不了解。 大兴区作为北京市的一个远郊区县，拥有着得天独厚的农业教育资源，在我们的身边就有向"大东农业观光园"这样的一个很好的农业教育基地。在大东农业观光园里面充分体现了现代农业。它目前拥有现代化日光温室34个，塑料大棚34个，露地80亩，为观光采摘、实验、示范、生产提供了场所。农业观光园里引进福建有机茶叶树，通过土壤改质能使其在北方正常生长；蔬菜嫁接的试验技术；大棚中的科学技术，如：所有地面具备节能、日光温室跨度大、内设供热、微喷、保湿、补光等多项功能；另外园区内设有科普知识长廊，为学生了解我国农业的历史文化和发展，创造了很好的条件。此基地实践活动适合小学高年级学生，他们可以在基地里面，进行参观、体验、学习了解高科技农业技术。通过基地的实践活动，可以增长学生现代农业知识，提高劳动能力，体验劳作的艰辛，培养学生良好的道德品质。可以很好的促进学生学习方式的变革，使学生直接从事实践性主题活动，实现学生自主学习和直接体验为主的学习方式，从而不