Week 6: Extended BVC Model using BCM Rule

Question 1

But BVC and sharp model can not account for long-term dynamics of place fields in similar enviroment

Answer 1

nor the importance of synaptic plasticity in place field stability

Question 2

Recall: Global remapping is not the whole story as there is long-term dynamics in place field cells - (4)

Answer 2

Global remapping is akin to change between attractor

But slow, experience dependent changes to place fields

As the attractor is slowly deformed over time (i.e. changing the active representation of one enviroment by moving to new enviroments)

Likely reflects synaptic plasticity and can not be explaiend by a ‘fixed’ model (lead to extended BVC model…)

Question 3

Extended BVC model where we added

Answer 3

learning rule called Bienestock-Cooper Monroe (BCM) rule

Question 4

BCM rule is a modification of

Answer 4

Hebbian plasticity

Question 5

In extended BVC model we update weights using BCM rule where (3)

Answer 5

• Requires pre and post-synaptic activity (xi and yi)
• Reduces if post-synaptic activity low
• Increases if activity is high

Question 6

In extended BVC model using BCM rule we change weight and update activity of yi (place cells)

Answer 6

sum of inputs xj (BVC) * weights and put through transfer function

Question 7

There is a sliding threshold in extended BVC model using

Answer 7

BCM rule

Question 8

The threshold of yi output firing in extended BVC model using BCM rule is

Answer 8

proportional to the square of recent activity of yi (Place cells)

Question 9

BCM rule explained is that you get pattern of activity x drives a neuron y (3)

Answer 9

y < threshold = weight decreases (unlearn weak inputs)
y > threshold = weight increases (learning)
Updating the threshold based on average activity of y

Question 10

Applying BCM rule to BVC (3)

Answer 10

• From BVCs, and get PC field representation
• Repeated so get PC field representation for all PC
• Update the weights between BVC and PC using BCM rule
• What happens if you have a weak secondary peak in your PC at a specific location in enviroment then learning rule will weaken the weights for the neuron that gives you this weak secondary peak over time

Question 11

UnSupervised learning (don’t impose any teaching rules) examples (3)

Answer 11

• Hebbian learning
• Competitive learning rule
• BCM Rule