




版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)
文檔簡介
1、AI & NN Lecture Notes Chapter 10 New Progress in Intelligent Systems 10.1 Basic Review1. Fundamental Features of BNN(1) Low speed of neuronal operation: 10(2) Huge number of neurons: 10 - 10(3) Highly interconnected: 10 - 10 inputs for each neuron leading to highly parallel operation and statist
2、ically and collectively decision-making(4) Learning by connection-strength revision(5) Binary communication among neurons while analog representation within neuron(6) Output information is continuous variable- 3 Hz111234(7) Connection structure observes constraints topologically and geometrically(8)
3、 Distributed control and integrative operation(9) Proper degeneration: system performance will gradually be degraded as number of unproper neurons increases; but there does not exist a critical point at which the system performance dies out totally(10) Relaxation is the basic computation manner: by
4、iteration, the system gradually satisfies a great number of weak constraints1. Fundamental Features of BNN (Cont.)2. ANN Features(1) Large scale complex system - Many parameters adjustable(2) Parallel processing - Very rapid processing(3) Distributes storage - Associative memory and holograph(4) Fle
5、xible configuration - Better adaptive(5) High redundancy - High robustness2. ANN Feature (Cont.)(6) Non-linear operation - Self-organization, chaos, and chaos synchronous(7) Collective computation - Powerful learning(8) Non-algorithmic - Good for experimental learning(9) Simple units - Easy to imple
6、ment by VLSI3. Biological Facts not Realized in ANN(1) Some neurons are to send excitatory signals and others inhibitory, yet artificial neurons are assumed to act in both possibilities.(2) The output peak (spike) of a biological neuron has certain computational significance, but artificial neurons
7、are designed to output an average.(3) The communications among biological neurons are in diffuse manner, but in artificial neuron it is designed in a point-to-point manner.3. Biological Facts Not Realized in ANN (Cont.)(4) It was discovered recently that both the time instant and the velocity of the
8、 pulses produced by neurons are meaningful to encoding, but only the information on velocity is utilized in ANN(5) There are over 200 kinds of biological neurons, nevertheless ANN has only one kind neuron.(6) There are many different kinds of neural media biologically, however, this fact is complete
9、ly ignored in ANN.4. Comparison between ES, NN and HumansItems Man ES ANN ComputerNumeric Calculation 4 3 2 1Operation Speed 4 3 1 2Computation Accuracy 4 2 3 1Knowledge Acquisition 1 3 2 4Knowledge Expression 1 3 2 4Parallel Processing 1 3 2 4Shallow Reasoning 1 3 2 4Deep Reasoning 1 3 2 4Items Man
10、 ES ANN ComputerImprecision Reasoning 1 2 3 4Heuristic Reasoning 1 2 3 4Learning Ability 1 3 2 4Stability 1 2 4 3Robustness 1 3 2 4Mesa Effect 1 4 2 3Creative Thinking 1 4 4 44. Comparison (Cont.) 10.2 New Progress in Depth: The Trinity Theory ofInformation, Knowledge & Intelligence1. The Model
11、of Intelligence AcquisitionExecution GoalIntelligence InformationKnowledgeCognitionDecision-making Problem/Environment2. A Simplified Model of IntelligenceCognitionInformationKnowledgeIntelligenceStrategyProducingInformation RefiningKnowledge Activation3. The State-of-The-Art of Intelligence Studies
12、 Information TheoryIntelligence TheoryKnowledge TheoryC. Shannon 1948M. Minsky etal 1956?Some of The References (1) BUPT Press., 1996 Comprehensive Information Theory - Syntactic Information: Form - Semantic Information: Meaning - Pragmatic Information: UtilitySome of The References (2) PTT Press, 1
13、992The slogan “AI is dead and long live NeuralNetworks” is totally wrong. The Intelligence Theory is the entirety of AI and ANN.- The 1st CNNC in 1990 in Beijing.Some of the References (3)“The Framework of Knowledge Theory” Chinese Engineering Science, No.10, 2000also ICSCI2000, Orlando, Florida, US
14、A-Definition & Categories of Knowledge: Quality-Measurements of Knowledge: Quantity-Reasoning of Knowledge: Content-Utility of Knowledge: Value -Knowledge & Information: Knowledge Formation-Knowledge & Intelligence: Knowledge Activation4. An Approach to Intelligence CreationS-I-RP-I-RF-K
15、-FV-K-FC-K-GI-S-GPROBLEMSe-I-GInformation and Knowledge BasesConsciousness: The Kernel of IntelligenceThe object with the form has such value.Form ConsciousnessValue Consciousness Content ConsciousnessStimulusRecognitionFormValueAssessmentMeaningProducingInductive Learning and Neural NetworksWTraini
16、ngSamplesDifferenceActual OutputDesired OutputRevision Strategy10.3 New Progress in Width AI & NN Converge toMAS/DAI & Intelligent Info-NetGeneral Model of MAS/DAICACACAPAPAMAPAPAMAPAPAMAResourcesP: ProcessingC: CommunicationM: ManagingFeatures of MAS/DAI1, All databases are physically distr
17、ibuted and neither global database and nor central control.2, All agents are physically separated and linked through communication network.3, Each agent is intelligent & autonomous.4, Agents work in parallel way.5, Agents are able to cooperate and to be coordinated.6, Agents share the resources
18、available. - AI ? ANN ? after all.Large Scale Intelligent Info-NetCom-NetAgentAgentAgentFunctionalities of AgentSensingTransferringCognition & DecisionTransferringControllingTaskSocial MemberSocial ToolNew Generation ToolProductive Task 10.4 Some of The Current Trends in Intelligence ResearchInternet-Oriented Intelligent Systems1. Intelligent Search Engine2. Intelligent Article Abstraction3. High-Quality Text Translation4. Text-Voice Conversion5. Multi-Agent Cooperative SystemAll have been carried on very promisingly in Research Center of Intell
溫馨提示
- 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
- 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
- 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
- 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
- 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。
最新文檔
- 醫(yī)療行業(yè)中的倫理決策支持系統(tǒng)建設(shè)
- 醫(yī)療人工智能技術(shù)在辦公健康管理中的應(yīng)用
- 全球化視角下跨國公司對(duì)智能醫(yī)療服務(wù)責(zé)任的考量
- 潁上縣交通運(yùn)輸局年春運(yùn)工作總結(jié)模版
- 以客戶為中心企業(yè)如何利用區(qū)塊鏈優(yōu)化客戶服務(wù)體驗(yàn)
- 兄弟分房合同范例
- 醫(yī)療大數(shù)據(jù)庫建設(shè)與健康管理的未來趨勢
- 語文《愛蓮說》課件
- 化學(xué)燒傷的臨床護(hù)理
- 溫州市普通高中2025屆高三第三次適應(yīng)性考試數(shù)學(xué)試題及答案
- 2025屆上海市華東師范大學(xué)二附中高三(最后沖刺)歷史試卷含解析
- 霧化吸入療法合理用藥專家共識(shí)(2024版)解讀 2
- (二模)2025年4月濰坊市高三高考模擬考試語文試卷(含答案)
- 單位食堂勞務(wù)外包服務(wù)投標(biāo)方案(技術(shù)方案)
- 養(yǎng)老院老年公寓養(yǎng)老機(jī)構(gòu)醫(yī)養(yǎng)結(jié)合機(jī)構(gòu)院科兩級(jí)質(zhì)量管理體系及目標(biāo)及實(shí)施方案2-5-10
- 貴州烏江煤層氣勘探開發(fā)有限公司招聘筆試題庫2025
- 穿線分包合同協(xié)議
- 叉車轉(zhuǎn)正考試題及答案
- 2025年磁化水防垢除垢裝置項(xiàng)目可行性研究報(bào)告
- 2024年司法考試歷年真題答案
- 2025年廣東省汕頭市高三二模生物試卷(含答案)
評(píng)論
0/150
提交評(píng)論