CLPS 0020- Lectures - Computations and Modeling

Question 1

What is the role of computational modeling?

Answer 1

requires quantifying, being explicit about ht enature of representations and processes; allows us ot test it and lesion it

Question 2

What is a symbolic approach to the mind?

Answer 2

cognitivism, build models of mind, takes computer metaphor literally

Question 3

What is an algorithm?

Answer 3

When the machinery doesn’t matter; structure doesn’t matter as long as the processes are the same

Question 4

What is intelligence?

Answer 4

Thinking rationally, following rules/algorithms, etc

Question 5

What is a Turing machine?

Answer 5

a hypothetical device that can perform any computational process that can be performed so long as it’s solved a set of rules/instructions with algorithms

Question 6

Do the symbols used in a Turing machine have any intrinsic meaning?

Answer 6

No: arbitrary

Question 7

What is deductive inference?

Answer 7

if the premises are true, then the conclusion is true

Question 8

What is inductive inference?

Answer 8

sample, then make a leap: the conclusion isn’t certain; includes probabilites; often used by scientific research

Question 9

What isn’t our reasoning completely logical?

Answer 9

affected by beliefs, knowledge, biases, etc.

Question 10

What is the importance of patterns in neural network models?

Answer 10

Very important!

Question 11

What are the correlates to neurons and synapses in the connectionist model?

Answer 11

Nodes and weights

Question 12

What does the architecture of models allow for in the connectionist model?

Answer 12

activaiton and inhibition among inputs

Question 13

What are representations in the connectionist model?

Answer 13

patterns of activation among the network’s elements

Question 14

What are representations in the symbolic model?

Answer 14

abstract symbols

Question 15

What is one of the major benefits to the connectionist model?

Answer 15

Helps explain the graded/probabilistic nature of representations

Question 16

What is the parallel distributed processing model?

Answer 16

All neurons act at once; no single neuron is responsible for any particular function; and the system is not just a simple memory device but uses info to go from a particular input to a particular output

Question 17

What is “distributed” about the parallel distributed processing model?

Answer 17

There’s no ear, nose, tail neuron for a dog representation: the combination of all of them collecitvely represents a dog as a whole

Question 18

List some properties of neural models.

Answer 18

Captures associations, correlations; is a feedback system; errors appear systematic; system may be interactive (input flow goes in multiple direcitons); probabilistc responses; graceful degradation; emergent properties

Question 19

What is a probabilistic model?

Answer 19

A combination of the symbolic and connectionist models

Question 20

What are the capabilities of probabilistic models?

Answer 20

Can infer definite answers, can infer graded/uncertain/probabilistic answers; inference in based on observations, which we do ALL the time in real life

Question 21

What is the Bayesian solution in probabilistic models?

Answer 21

probabilistic inference; narrow down to only things that could possibly hav ecaused the imaged, weights the various options according ot how likely they are, and chooses the one that fits best

Question 22

What is one of the main issues with the probabilistic model?

Answer 22

That is only operates on a computational level; characterizes what the mind is trying to accomplish, not HOW or with what materials