Neural Networks part 1

Question 1

what are the goals of science, quantitative models

Answer 1

Description, prediction, explanation

Question 2

what is retrograde?

Answer 2

It is a movement of a planet off its normal path. going around a point

Question 3

ptolemaic geocentric model allowed for what?

Answer 3

Predictions

but lacks satisfying explanation

Question 4

Copernican heliocentric model

Answer 4

First to say that we are actually going around the sun, interestingly the theory was worse at predicting

Question 5

Kepler’s law of planetary motion

Answer 5

allowed for accurate prediction and explanation

Question 6

why do we have quantitative models?

Answer 6

data never speaks for itself but requires a model to understand

verbal theorizing alone cannot substitute for quantitative analysis

always alternative models

model comparison rests on both quantitative evaluation and intellectual and scholarly judgment

Question 7

what is the fundamental tradeoff in models?

Answer 7

simplicity and elegance vs. complexity and accuracy

Goal: maximize explanatory power while minimizing complexity

Question 8

what makes a good model?

Answer 8

That there is a trade off between accuracy and simplicity

Question 9

what is the goal for an explination

Answer 9

explain as much as possible as simply as possible

Question 10

types of cognitive models

Answer 10

Mathematical model (Fitt’s law)

Symbolic models
-describes what goes on in the mind as a symbolic representation

dynamical systems models
-think of the mind as a point moving through an abstract mental state space

Hybrid models

Question 11

computational neuroscience

Answer 11

How does a single neuron network

modelling a single neuron in lots of detain

Question 12

Artificial neural network:

Current state

Answer 12

Each unit has an activation which changes rapidly from moment to moment based on current input
-eg neurons rapidly firing action potentials

Question 13

Learned information

Answer 13

Each connection has a weight which changes slowly based on learning
-eg synaptic strength changing slowly die to experience

Question 14

Topology

Answer 14

Feed forward, simple recurrent (elman), self organizing map (Kohonen), Fully recurrent

Question 15

simple recurrent

Answer 15

Adding a loop not just feed forward

Question 16

Self organizing map

Answer 16

A lattice where each one is connected to the neighbour

Question 17

Fully recurrent

Answer 17

Everything is connected to everything and information is flowing freely

Question 18

learning:

Supervised learning

Answer 18

Adjust weights based on difference between actual output and correct output

when you are wrong you get given the correct answer

Question 19

unsupervised learning

Answer 19

Adjust weights based on correlations in input “neurons that fire together wire together”

what things go together

Question 20

reinforcement learning

Answer 20

Adjust weights based on difference between actual reward and expected reward

Question 21

how do you adjust the weights in supervised feedforward networks

Answer 21

Back propagation

stats at output layer and work our way back

Question 22

unsupervised recurrent (hopfield) networks

Answer 22

each neuron is connected to every other neuron. a total connection in the network

Question 23

what is the use of unsupervised recurrent (hopfield) networks

Answer 23

gives association to memory

store pattens

in future if you give it a similar cue/pattern it will recover the original

works a lot like out long term memory

Question 24

Unsupervised learning

Answer 24

expose the network to an image and it just re wires itself so that image is represented in the network and it uses Hebb’s ruel

Question 25

What is Hebb's rule?

Answer 25

Neurons that fire together, wire together, neurons that fire out of sync, fail to link.

Question 26

Self organizing Maps (kohonen)

Answer 26

net works that receive multi dimensional input you give it different inputs and organizes the space of possible inputs into a map

Question 27

Output for kohonen map

Answer 27

Chose the cell whose weight vector is the nearest too the input

Question 28

Updating for Kohonen map

Answer 28

Updating each cell's weight vector to be more like the input vector the amount of updating decreases with 1) topological distance from the chosen cel and 2) time

Question 29

Supervised recurrent (Elman) networks

Answer 29

look like feed forward but add in a context layer output to hidden then feeds it back to context

Question 30

What does it give the network?

Answer 30

interesting memory normally with feedforward it goes with input but with this it allows it to to keep a record and memory ( remember what it has seen in recent past) -allows it to remember sequences through time

Question 31

what is the simple recurrent network good for?

Answer 31

Remembering sequences throughout time

Question 32

what is the simple recurrent network good at predicting

Answer 32

What letter will come next what letter is likely to follow other letters

Question 33

as you get further into a word simple recurrent network has a better change of what?

Answer 33

Guessing the next letter similar to how kids learn language

Question 34

Reinforcement learning

Answer 34

you over time what actions lead to reward and do that more ofter (maximize)

Question 35

what is the RL problem

Answer 35

Learn what action to take to maximize total reward given the surrent state of the environment no one is teaching you but you are learning from feedback

Question 36

what does deep mean? RL

Answer 36

That there are many layers - this is helpful because with each layer the network can learn things at a slightly more abstract level