Reinforcement

Question 1

If the probability of choosing action i is:
p_i=(exp(ßm_i)/(sum(ßm_j))
What are the variables?

Answer 1

m_i is the estimated reward for action i.
ß is the choosiness or exploitation parameter.
If ß=0 then all actions are chosen with equal probability and there is little discrimination; if ß is large the action with the heighest estimated reward is chosen with probability near one, there is little exploration.

Question 2

How is m_i updated?

Answer 2

m_i-> m_i + eta∂
where eta is the learning rate
∂=r-m_i

Question 3

What is the basal ganglia and where is it located?

Answer 3

The basal ganglia is a collection of sub-cortical brain areas or nuceli found near the centre of the brain and connected to the cortex, thalmus and brain stem.

Question 4

What three parts of the brain is the basal ganglia connected to?

Answer 4

Thalmus, cortex, brain stem.

Question 5

What is thought to be the function of the Basal ganglia?

Answer 5

Important in decision making, action selection and in the regulation of some routine behaviours like eye movements.