Presentation is loading. Please wait.

Presentation is loading. Please wait.

Tom Griffiths CogSci C131/Psych C123 Computational Models of Cognition.

Similar presentations


Presentation on theme: "Tom Griffiths CogSci C131/Psych C123 Computational Models of Cognition."— Presentation transcript:

1 Tom Griffiths CogSci C131/Psych C123 Computational Models of Cognition

2 Tom Griffiths CogSci C131/Psych C123 Computational Models of Cognition

3 ComputationCognition

4 Cognitive science The study of intelligent systems Cognition as information processing inputoutputinputoutput

5 computation Computational modeling Look for principles that characterize both computation and cognition inputoutputinputoutput

6 Two goals Cognition: –explain human cognition (and behavior) in terms of the underlying computation Computation: –gain insight into how to solve some challenging computational problems

7 Computational problems Easy: –arithmetic, algebra, chess Difficult: –learning and using language –sophisticated senses: vision, hearing –similarity and categorization –representing the structure of the world –scientific investigation human cognition sets the standard

8 Three approaches Rules and symbols Networks, features, and spaces Probability and statistics

9 Three approaches Rules and symbols Networks, features, and spaces Probability and statistics

10 Logic All As are Bs All Bs are Cs All As are Cs Aristotle (384-322 BC)

11 Mechanical reasoning (1232-1315)

12 The mathematics of reason Thomas Hobbes (1588-1679) Rene Descartes (1596-1650) Gottfried Leibniz (1646-1716)

13 Modern logic George Boole (1816-1854) Friedrich Frege (1848-1925) PQPQPQPQ

14 Computation Alan Turing (1912-1954)

15 The World Logic P  Q P Q Facts Inference Rules

16 Categorization

17 cat  small  furry  domestic  carnivore

18 The World Logic P  Q P Q Facts Inference Rules

19 The World Early AI systems… Workspace Operations Actions Observations Facts Goals Operations

20 Rules and symbols Perhaps we can consider thought a set of rules, applied to symbols… –generating infinite possibilities with finite means –characterizing cognition as a “formal system” This idea was applied to: –deductive reasoning (logic) –language (generative grammar) –problem solving and action (production systems)

21 Language as a formal system Noam Chomsky

22 Language “a set (finite or infinite) of sentences, each finite in length and constructed out of a finite set of elements” all sequences L This is a good sentence1 Sentence bad this is0 linguistic analysis aims to separate the grammatical sequences which are sentences of L from the ungrammatical sequences which are not

23 A context free grammar S  NP VP NP  T N VP  V NP T  the N  man, ball, … V  hit, took, … S NP VP T N V NP T N the man hit the ball

24 Rules and symbols Perhaps we can consider thought a set of rules, applied to symbols… –generating infinite possibilities with finite means –characterizing cognition as a “formal system” This idea was applied to: –deductive reasoning (logic) –language (generative grammar) –problem solving and action (production systems) Big question: what are the rules of cognition?

25 Computational problems Easy: –arithmetic, algebra, chess Difficult: –learning and using language –sophisticated senses: vision, hearing –similarity and categorization –representing the structure of the world –scientific investigation human cognition sets the standard

26 Inductive problems Drawing conclusions that are not fully justified by the available data –e.g. detective work Much more challenging than deduction! “In solving a problem of this sort, the grand thing is to be able to reason backward. That is a very useful accomplishment, and a very easy one, but people do not practice it much.”

27 Challenges for symbolic approaches Learning systems of rules and symbols is hard! –some people who think of human cognition in these terms end up arguing against learning…

28 The poverty of the stimulus The rules and principles that constitute the mature system of knowledge of language are actually very complicated There isn’t enough evidence to identify these principles in the data available to children Therefore Acquisition of these rules and principles must be a consequence of the genetically determined structure of the language faculty

29 The poverty of the stimulus Learning language requires strong constraints on the set of possible languages These constraints are “Universal Grammar”

30 Challenges for symbolic approaches Learning systems of rules and symbols is hard! –some people who think of human cognition in these terms end up arguing against learning… Many human concepts have fuzzy boundaries –notions of similarity and typicality are hard to reconcile with binary rules Solving inductive problems requires dealing with uncertainty and partial knowledge

31 Three approaches Rules and symbols Networks, features, and spaces Probability and statistics

32 Similarity What determines similarity?

33 Representations What kind of representations are used by the human mind?

34 Semantic networksSemantic spaces Representations How can we capture the meaning of words?

35 Categorization

36 Computing with spaces x 1 x 2 y perceptual features +1 = cat, -1 = dog x1x1 x2x2 y dog cat error:

37 Networks, features, and spaces Artificial neural networks can represent any continuous function…

38 Problems with simple networks x 1 x 2 x1x1 x2x2 y Some kinds of data are not linearly separable x1x1 x2x2 AND x1x1 x2x2 OR x1x1 x2x2 XOR

39 A solution: multiple layers z 1 z 2 y x 1 x 2 y z1z1 z2z2 x1x1 x2x2 hidden layer input layer output layer

40 Networks, features, and spaces Artificial neural networks can represent any continuous function… Simple algorithms for learning from data –fuzzy boundaries –effects of typicality

41 E (error) w ij (  is learning rate) General-purpose learning mechanisms

42 The Delta Rule x 1 x 2 y +1 = cat, -1 = dog output error influence of input for any function g with derivative g perceptual features

43 Networks, features, and spaces Artificial neural networks can represent any continuous function… Simple algorithms for learning from data –fuzzy boundaries –effects of typicality A way to explain how people could learn things that look like rules and symbols…

44 Simple recurrent networks z 1 z 2 x 1 x 2 hidden layer input layer output layer context units input (Elman, 1990) x 2 x 1 copy

45 Hidden unit activations after 6 iterations of 27,500 words (Elman, 1990)

46 Networks, features, and spaces Artificial neural networks can represent any continuous function… Simple algorithms for learning from data –fuzzy boundaries –effects of typicality A way to explain how people could learn things that look like rules and symbols… Big question: how much of cognition can be explained by the input data?

47 Challenges for neural networks Being able to learn anything can make it harder to learn specific things –this is the “bias-variance tradeoff”

48 Bias-variance tradeoff

49

50

51

52 What about generalization?

53 What happened? The set of 8th degree polynomials contains almost all functions through 10 points Our data are some true function, plus noise Fitting the noise gives us the wrong function This is called overfitting –while it has low bias, this class of functions results in an algorithm that has high variance (i.e. is strongly affected by the observed data)

54 The moral General purpose learning mechanisms do not work well with small amounts of data (the most flexible algorithm isn’t always the best) To make good predictions from small amounts of data, you need algorithms with bias that matches the problem being solved This suggests a different approach to studying induction… –(what people learn as n  0, rather than n   )

55 Challenges for neural networks Being able to learn anything can make it harder to learn specific things –this is the “bias-variance tradeoff” Neural networks allow us to encode constraints on learning in terms of neurons, weights, and architecture, but is this always the right language?

56 Three approaches Rules and symbols Networks, features, and spaces Probability and statistics

57 Probability Gerolamo Cardano (1501-1576)

58 Probability Thomas Bayes (1701-1763) Pierre-Simon Laplace (1749-1827)

59 Bayes’ rule Posterior probability LikelihoodPrior probability Sum over space of hypotheses h: hypothesis d: data How rational agents should update their beliefs in the light of data

60 Cognition as statistical inference Bayes’ theorem tells us how to combine prior knowledge with data –a different language for describing the constraints on human inductive inference

61 Prior over functions k = 8,  = 5,  = 1 k = 8,  = 5,  = 0.1 k = 8,  = 5,  = 0.3 k = 8,  = 5,  = 0.01

62 Maximum a posteriori (MAP) estimation

63 Cognition as statistical inference Bayes’ theorem tells us how to combine prior knowledge with data –a different language for describing the constraints on human inductive inference Probabilistic approaches also help to describe learning

64 Probabilistic context free grammars S  NP VP1.0 NP  T N0.7 NP  N 0.3 VP  V NP1.0 T  the0.8 T  a0.2 N  man0.5 N  ball0.5 V  hit0.6 V  took0.4 S NP VP 1.0 T N 0.7 V NP 1.0 the 0.8 man 0.5 hit 0.6 the 0.8 ball 0.5 T N 0.7 P(tree) = 1.0  0.7  1.0  0.8  0.5  0.6  0.7  0.8  0.5

65 Probability and learnability Any probabilistic context free grammar can be learned from a sample from that grammar as the sample size becomes infinite Priors trade off with the amount of data that needs to be seen to believe a hypothesis

66 Cognition as statistical inference Bayes’ theorem tells us how to combine prior knowledge with data –a language for describing the constraints on human inductive inference Probabilistic approaches also help to describe learning Big question: what do the constraints on human inductive inference look like?

67 Challenges for probabilistic approaches Computing probabilities is hard… how could brains possibly do that? How well do the “rational” solutions from probability theory describe how people think in everyday life?

68 Three approaches Rules and symbols Networks, features, and spaces Probability and statistics

69


Download ppt "Tom Griffiths CogSci C131/Psych C123 Computational Models of Cognition."

Similar presentations


Ads by Google