The Connectionist Approach

Question 1

Connectionism

Answer 1

Approach to creating computer models for representing cognitive processes

Also called “parallel distributed processing” models because they propose concepts that are represented in the distributed activity of many linked units

Question 2

Input units

Answer 2

Activated by stimulation from the environment

Question 3

Hidden units

Answer 3

Receive input from input units

Question 4

Output units

Answer 4

Receive input from hidden units

Question 5

Connection weights

Answer 5

Determine how strongly signals from one unit increase or decrease activity of next unit

Question 6

How do connectionist networks learn?

Answer 6

Don’t have knowledge programmed in
They begin with equal or random response parameters, then the network is trained over many trials
If a mistake is made, error signal is generated

Question 7

back-propagation

Answer 7

Process wherein error signal transmitted back through circuit

Indicated how connection weights should be changed to allow the output signal to match the correct signal, process repeats until error is zero

Question 8

graceful degradation

Answer 8

Disruption of performance occurs gradually as parts of the system are damaged

Question 9

Advantages

Answer 9

Success in simulating cognitive processes

Can explain generalization of learning (similar concepts will have similar connectionist patterns)

Seems analogous to real brains/neurons