Connectionism and Classical Conditioning

Connectionism and Classical Conditioning

Connectionism and Classical Conditioning Michael R.W. Dawson University of Alberta Edmonton, Alberta Canada T6G 2P9 1-(780)-492-5175 Abstract The purpose of this monograph is to examine the relationship between a par- ticular artificial neural network, the perceptron, and the Rescorla-Wagner model of learning. It is shown that in spite of the fact that there is a formal equivalence be- tween the two, they can make different predictions about the outcomes of a num- ber of classical conditioning experiments. It is argued that this is due to algo- rithmic differences between the two, differences which are separate from their computational equivalence. [email protected] Rosenblatt Software Available As Freeware At: http://www.bcp.psych.ualberta.ca/~mike/Software/Rosenblatt/index.html A peer-reviewed monograph published by Comparative Cognition and Behavior Reviews on behalf of the Comparative Cognition Society. © Michael R. W. Dawson 2008 Chapter 1: Learning in Networks and Animals_____________________________ 9 1.1 An Empirical Puzzle __________________________________________________ 10 1.1.1 Identical, But Different _____________________________________________________ 10 1.1.2 Networks and Learning_____________________________________________________ 10 1.2 A Formal Surprise____________________________________________________ 11 1.2.1 Perceptron Nonlinearity ____________________________________________________ 11 1.2.2 A Linear Proof ___________________________________________________________ 11 1.2.3 A Nonlinear Proof_________________________________________________________ 11 1.3 Cognitive Science and Associative Learning_______________________________ 12 1.3.1 Cognition vs. Association ___________________________________________________ 12 1.3.2 Levels and Equivalence ____________________________________________________ 12 1.4 The Synthetic Approach _______________________________________________ 13 1.4.1 The Analytic Approach_____________________________________________________ 13 1.4.2 A Synthetic Alternative_____________________________________________________ 13 1.4.3 Synthetic Association ______________________________________________________ 13 1.5 Purposes and Intended Readership ______________________________________ 14 1.5.1 Purposes ________________________________________________________________ 14 1.5.2 Intended Readership _______________________________________________________ 14 1.6 What Is This Book Not About? _________________________________________ 15 1.6.1 Multi-layer Perceptrons ____________________________________________________ 15 1.6.2 Temporal Models _________________________________________________________ 15 1.6.3 Rescorla-Wagner Alternatives _______________________________________________ 15 1.7 What Is This Book About? _____________________________________________ 16 1.7.1 What Lies Ahead _________________________________________________________ 16 1.7.2. What Lay Behind_________________________________________________________ 16 Chapter 2: The Perceptron ____________________________________________ 17 2.1 Neuronal Inspiration __________________________________________________ 18 2.1.1 Functional Explanations ____________________________________________________ 18 2.1.2 Neuronal Function ________________________________________________________ 18 2.2 A Digital Output Unit _________________________________________________ 19 2.2.1 Perceptron as Neuron ______________________________________________________ 19 2.2.2 Computing Net Input ______________________________________________________ 19 2.2.3 Converting Net Input to Activity _____________________________________________ 19 2.3 Activation and Response_______________________________________________ 20 2.3.1 What Is Activity? _________________________________________________________ 20 2.3.2 Activity and Response _____________________________________________________ 20 2.4 Association in the Perceptron___________________________________________ 21 2.4.1 Stimulus and Response _____________________________________________________ 21 2.4.2 UCS→UCR in the Perceptron _______________________________________________ 21 2.5 Multiple Associations _________________________________________________ 22 2.5.1 Multiple Inputs and Outputs _________________________________________________ 22 2.5.2 An Example Perceptron ____________________________________________________ 22 2.6 Learning in Perceptrons _______________________________________________ 23 2.6.1 Neurons That Learn _______________________________________________________ 23 2.6.2 Supervised Learning _______________________________________________________ 23 A peer-reviewed monograph published by Comparative Cognition and Behavior Reviews on behalf of the Comparative Cognition Society. © Michael R. W. Dawson 2008 2.7 Hebb Learning _______________________________________________________ 24 2.7.1 The Law of Exercise_______________________________________________________ 24 2.7.2 The Hebb Rule ___________________________________________________________ 24 2.8 Pros and Cons of Hebb Learning________________________________________ 25 2.8.1 Hebb Rule Advantages _____________________________________________________ 25 2.8.2 Hebb Rule Disadvantages___________________________________________________ 25 2.8.3 Learning from Mistakes ____________________________________________________ 25 2.9 The Delta Rule _______________________________________________________ 26 2.9.1 Defining Error____________________________________________________________ 26 2.9.2 Delta Learning Of Associations ______________________________________________ 26 2.9.3 Supervised Learning and US_________________________________________________ 26 Chapter 3: Associative Learning in the Perceptron: Case Studies _____________ 27 3.1 Studying Associative Learning in the Perceptron __________________________ 28 3.1.1 Setting the Stage __________________________________________________________ 28 3.1.2 Technical Details _________________________________________________________ 28 3.2 Simple Association____________________________________________________ 29 3.2.1 Classical Conditioning _____________________________________________________ 29 3.2.2 Training Set _____________________________________________________________ 29 3.2.3 Results__________________________________________________________________ 29 3.3 Intensity of Conditioned Stimuli ________________________________________ 30 3.3.1 Conditioning and CS Intensity _______________________________________________ 30 3.3.2 CS Intensity in the Perceptron _______________________________________________ 30 3.3.3 Results__________________________________________________________________ 30 3.4 Intensity of Unconditioned Stimuli ______________________________________ 31 3.4.1 The Effect of US Intensity __________________________________________________ 31 3.4.2 Effects of Decreasing η ____________________________________________________ 31 3.4.2 Effects of Increasing η _____________________________________________________ 31 3.5 Configural Representations of Compound Stimuli _________________________ 32 3.5.1 Compound Stimuli ________________________________________________________ 32 3.5.2 Elemental Representations __________________________________________________ 32 3.5.2 Configural Representation __________________________________________________ 32 3.6 Positive Patterning In The Perceptron ___________________________________ 33 3.6.1 Positive Patterning ________________________________________________________ 33 3.6.2 Training Set _____________________________________________________________ 33 3.6.3 Results__________________________________________________________________ 33 3.6.4 The Modern Perceptron ____________________________________________________ 33 Chapter 4: Modernizing the Perceptron: The Integration Device _____________ 34 4.1 A Continuous Approximation of the Heaviside Equation ____________________ 35 4.1.1 From Discrete To Continuous________________________________________________ 35 4.1.2 The Logistic Activation Function _____________________________________________ 35 4.2 Training an Integration Device _________________________________________ 36 4.2.1 The Delta Rule ___________________________________________________________ 36 4.3 Acquisition Curves for the Integration Device _____________________________ 37 4.3.1 General Method __________________________________________________________ 37 4.3.2 Association and Extinction __________________________________________________ 37 4.3.2 Results__________________________________________________________________ 37 4.4 On Two Responses____________________________________________________ 38 A peer-reviewed monograph published by Comparative Cognition and Behavior Reviews on behalf of the Comparative Cognition Society. © Michael R. W. Dawson 2008 4.4.1 Go or No Go _____________________________________________________________ 38 4.4.2 Approach or Avoid ________________________________________________________ 38 4.5 The Hyperbolic Integration Device ______________________________________ 39 4.5.1 The Hyperbolic Tangent ____________________________________________________ 39 4.5.2 Hyperbolic Learning Rule___________________________________________________ 39 4.6 Approach or Avoid ___________________________________________________ 40 4.6.1 Inhibition and Excitation____________________________________________________ 40 4.6.2 Training Set _____________________________________________________________ 40 4.6.3 Results__________________________________________________________________

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    115 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us