东南大学 崇志宏:数据与智能暑期研讨班提纲

所需积分/C币:13 2017-07-16 09:37:41 1.6MB PDF

符号和联结是认知建模的基本方式,两者的有机融合是目前研究的趋势,是认知计算的基本方式,也是处理大数据的基本工具。
参考资料 际 te lab 数据&智能实验室 i Deep Learning ii Probabilistic Graphical Models Principles and Techniques an Yoshua Bengio daphne Koller L- Aaron Courville NIr Friedman Machine earning robabilistic Perspective ■PMrD量v 其他相关文献(PPT内标注 Raul Rojas Neural Networks IDEA A Systematic Introduction isual 邮wdea ·Mind imaginatio DESIGN M arkov Logic mind.CLE& solution Behaviour An Interface laver for image aRTCONCEPTION h MAkT A creative DESIGN Artificial intelli ence 对认知建模的可能方法 Domingos and Daniel Lowd Uanersty of Washington Seal 涉及的领域 际 te lab 数据&智能实验室 Economics Computer science and Animal learning Organizational (Cognitive science, Psychology, Behavior Neuroscience) Machine learning Adaptive Control Evolution Theory Statistics 东南大学数据与智能实验室D& ntel lab 问题解决:模型和计算 际 te lab 数学 数据&智能实验室 模型 Everything should be made as simple as possible, but not simpler. 洞察 效果 效率 抽象 客观 计算 世界 模型Ⅲ计算和工具 对认知建模的可能方法 东南大学数据与智能实验室D& ntel lab 符号和联结:认知建模的可能方法 Artificial Neural Networks 际 A symbol system Is The perceptron D'o ntel lab (1)a set of arbitrary physical tokens(scratches 数据智能实验室 on paper, holes on a tape, events in a digital Activation Fundamental unit of a Neural network function computer, etc. that are (2)manipulated on the basis of explicit rules that are (3)likewise physical tokens and strings of to lf∑wx> kens. The rule-governed symbol-token manipula l otherwise ion is based ∑ weights (4)purely on the shape of the symbol tokens (not their "meaning"), i.e. it is purely syr atacTIc. Annuae hidden layer 1 hidden layer 2 hidden layer 3 and consists of input layer (5) ruefully combining and recombining symbol tokens. There are output layer (6)primitive atomic symbol tokens and (7)composite symbol-token strings. The entire system and all its parts- the atomic tokens, the composite tokens, the syntactic manipulations (both actual and possible) and the rules -are all ( 8)semantically interpretable: The syntax can be systematically assigned a meaning(e.g. as standing for objects, as describing states of aftairs) 对认知建模的可能方法 THE SYMBOL GROUNDING PROBLEM How to grow a mind: statistics Structure and abstraction Stevan HARNAD Joshua B. Tenenbaum, .Charles Kemp, Thomas L. Griffiths, Noah D. Goodman" 符号和联结:认知建模的可能方法 Artificial Neural Networks 际 The perceptron D'o ntel lab 数据智能实验室 Towards Bayesian Deep learning Activation Fundamental unit of a Neural Network function A Survey maro logic li∑wx>0 ce Laver for Artificial intelligence l otherwise ∑ Pedro domingos and Daniel Lowd weights Hao Wang, Dit- Yan Yeung Uanersty of WashingtoN, Seattl AnniNe nidden layer 1 hidden layer 2 hidden layer 3 input layer Probabilistic Gral output layer THE SYMBOL GROUNDING PROBLEM The curse off Stevan HARNAD Dimentioality! 对高维不确定定、不完全、不一致建模的方法 THE SYMBOL GROUNDING PROBLEM How to grow a mind: statistics Structure and abstraction Stevan HARNAD Tenenbaum, Charles Kemp, Thomas L. Griffiths, Noah D. Goodma 主要技术方法 际 te lab 数据&智能实验室 I. Inference as Optimization Variational Reparameterization BP through Sampling Operation Mean-->Expectation: Sampling lI. Learning as Optimization NonConvex Optimization I. The Structure of Distribution& Density: z 东南大学数据与智能实验室D& ntel lab 符号和联结:融合的例子(贝叶斯模型) D ntel lab Examples 数据&智能实验室 16 0.5 4812162024283236404448525660646872768084889296100 60 0.5 4812162024283236404448525660646872768084889296100 168264 0.5 hhI 4812162024283236404448525660646872768084889296100 16231920 0.5 oH皿 4812162024283236404448525660646872768084889296100 东南大学数据与智能实验室D& ntel lab 符号和联结:融合的例子(贝叶斯模型) data 16 data=168264 eve odd even squares mult of 3 squares mult of 4 mult of 3 ult of 4 ult of5 ult of 5 mult of 6 mult of 6 mult of 7 mult of 7 mult of 8 mult of 8 mult of 9 mult of 9 mult of 10 mult of 10 ends in t ends in 1 ends in 2 ends in 2 ends in 3 ends in 3 ends in ends in 4 ends in 5 ends in 5 ends in 6 ends in 6 ends in 7 ends in 7 15 ends in 8 ends in 8 ends in 9 ends in 9 of 2 powers of powers of 3 powers of 3 powers of 10 powers of 10 powers of 5 powers of 5 wers of 6 powers of 6 powers of 7 ers rs of powers of 8 powers of 9 5 powers of 9 powers of 1 powers of 1 al al powers of 2 +(37 ers of 2+37 powers of 2-(32 powers of 2-(32 0.1 020 20 0.5 0.1 020 0.2 0.40 02 0.4 pnor x10 post pn lk post

...展开详情
上传资源赚积分,得勋章
最新资源