PSYC 532 Lecture Notes - Short-Term Memory, Spreading Activation, Connectionism

26 views4 pages
Connectionism Lecture Notes
These are additional notes to the slides so it would be best to follow along with the slides for full
coverage of the topic….
Each Unit Has Simple Program:
Difference between biology and computer= excitatory and inhibition
Psychological Equivalents:
Pattern of activation across units (active or not)
Long term v.s Short term memory
Connection weights are being adjusted
Adjustment of connection weights= in order to reduce error
~Most general scheme…
Auto-Associator network:
Circles= units which are all doing the same job + fully connected to each other
Inputs-all are receiving this
Outputs-all sending this
Units + connection weights
Taking out some unit will make it do different things
Other Networks Have Restrictions:
Some constraints on values of connections e.g: Some groups of units have mutually inhibitory
connections (which are competing with each other)
~Good at learning patterns + generalization…
Output units= answer to problem
Hidden units= internal computation, no output or input
Input units= describe particular problem working on
o No lateral connections
o No communicating within a layer
o No backward connections
Distributed representations:
Each rectangle represents a neuron
Pattern represents a concept or an idea
Each different pattern represents a different idea
Each unit (96) is participating in different concepts
Unlock document

This preview shows page 1 of the document.
Unlock all 4 pages and 3 million more documents.

Already have an account? Log in

Document Summary

These are additional notes to the slides so it would be best to follow along with the slides for full coverage of the topic . Difference between biology and computer= excitatory and inhibition. Pattern of activation across units (active or not) Adjustment of connection weights= in order to reduce error. Circles= units which are all doing the same job + fully connected to each other. Taking out some unit will make it do different things. Some constraints on values of connections e. g: some groups of units have mutually inhibitory connections (which are competing with each other) Hidden units= internal computation, no output or input: no lateral connections, no communicating within a layer, no backward connections. Pattern represents a concept or an idea. Each different pattern represents a different idea. Each unit (96) is participating in different concepts. Each concept requires many units (can represent a large amount of concepts) Distributed= 96 = 2 (more efficient especially in ai)

Get access

Grade+
$10 USD/m
Billed $120 USD annually
Homework Help
Class Notes
Textbook Notes
40 Verified Answers
Study Guides
Booster Classes
Class+
$8 USD/m
Billed $96 USD annually
Homework Help
Class Notes
Textbook Notes
30 Verified Answers
Study Guides
Booster Classes