Notesale: Turn your study into money

Document Preview

Extracts from the notes are below, to see the PDF you'll receive please use the links above

---

15th November

Hippocampal and striatal roles in spatial navigation
Morris water maze (Morris et al
...
The rat swims around in opaque
water, trying to find a platform (below the
water level) to rest on
...
If the rat has a lesion that
damaged the hippocampus, then the rat
performs less well, compared to cortical
lesioned animals
...
The
hippocampus lesioned animals swim quite
randomly, they think that it’s near the
edge of the wall, but doesn’t know which
quadrant
...
1997) - translates place cell activity with simple Hebbian learning to produce
simple memory system
When at the goal (e
...
relief of
climbing onto the platform, or food)
some cells in other parts of the brain
codes for this reinforcement
...
Then Hebbian learning
(or unsupervised learning rule) causes
the synapses from the place cells that
happen to be active at the goal
location to be strengthened
...
This could be the simplest model for spatial memory, in order to go back to the place
where the rat found the chocolate, the rat monitors the firing rate of the goal cell, the rat just has to move
around so as to increase that firing rate of the goal cell to get back to the goal
...
This goal cell
provides an internal estimate of how well you are doing
...
This is similar to Simon and Barto’s reinforcement learning algorithm
...
It requires hunting, but doesn’t taken into account learning
...
g
...


15th November
This model can be consistent with goal-independent ‘latent learning’
...
The previous exploration tunes place cells, which helps the animal when it does get the
reward, as it will know where it is better
...

A neural circuit for spatial cognition in and around the hippocampus

Head-direction cells are found in the presubiculum and medial entorhinal cortex, mixed with the grid cells
...
Head direction cells are in
layer 2 and presubiculum
...

Striatum




The “Striatum” or “Basal ganglia”: caudate, putatmen, nucleus accumbens, globus pallidus
...

Important for generating movement, and planning movement
...

‘Place’ learning is dependent on hippocampus (e
...
Morris et al
...
g
...
Rats were trained to approach a
consistently baited arm in a cross-maze from the same start box (four trials/day/14 total days)
...
Three minutes prior to the probe trial, rats received bilateral injections

15th November
of either saline or a 2% lidocaine solution (in order to produce neural
inactivation) into either the dorsal hippocampus or dorsolateral caudate
nucleus
...
e
...
e
...

Saline-treated rats displayed place learning on the Day 8 probe trial and
response learning on the Day 16 probe trial, indicating that with extended
training there is a shift in learning mechanisms controlling behaviour
...

Rats given lidocaine injections into the caudate nucleus displayed place learning on both the Day 8 and the
Day 16 probe trials, indicating a blockade of response learning following inactivation of the caudate nucleus
...

Brown & Sharp (1995) model: learning to make correct moves from place and direction information
Inputs were place cells in the hippocampus
and head direction cells
Output cells were nucleus accumbens
(basal ganglia/motor output); these output
cells should say left/right/ahead
...
Then the connection weights will learn
to generate outputs of left, right or ahead
...
Navigation is non-linearly separable
...
Same goes for head direction cells, some will fire when
it’s facing north, some will fire when it’s facing south
...
So
the weight of the connection is strengthened
...
So half of the time the connection weight is
strengthened, half of the time, the weight is weakened
...
Each motor cell, when activated, causes a particular locomotor
movement in a simulated rat
...
The connection strengths between cells are set randomly at the start of a simulation, so
that the animal’s movements are initially random
...

Recently active synapses which led to the “correct” locomotor recently active synapses which led to the
“correct” locomotor response are strengthened, so that any move which took place in a particular locational
and directional context, and resulted in reinforcement, is “stamped in” in a Thorndikian (1898, 19 11)
manner
...

In the actual accumbens, this neuronal plasticity would be assumed to be initiated by the release of
dopamine, which would be caused by the receipt of reward in the designated reward location
...

The current model’s performance was tested in simulations of the Morris water maze task, and compared
against the performance of actual rats, as reported by Morris (1981)
...
The water is made opaque by the addition of milk, and animals
learn to navigate to a hidden platform, placed at a fixed location just below the water surface, in order to
escape
...

Sensory stimuli close to the animal in that location activate cells in the sensory cell layer, in a manner
described below
...
In
addition, a particular set of head direction cells are activated in relation to the animal’s initial directional
heading
...
This step places the animal
into a slightly different location, and the whole cycle, as just described, repeats
...
Activation of a single hippocampal cell uniquely blocks the output from all but two
motor units, one unit in each cluster
...

So the nucleus accumbens cells get a one to one connection with each place cell
...

So the difficult thing is to decide which pool of neurons should get more activation to decide whether the
animal turns left or right, and this is dependent on head direction cells
...


15th November
There needs to be a way of learning of the goal location to modify the head direction cells projecting onto
the nucleus accumbens cells, i
...
associating being north and turning left to get to the goal
...
So they calculated a recency weighted, cumulative track of Hebbian activity of the
connections from HD cells to NA cells
...
If we can keep track of this information, when we do
finally get to the goal, we can look back and say, recently which connections had Hebbian activity, then give
those connections the deserved ‘credit’ for taking us to the goal
...
This seems reasonable since actions taken most recently are those most closely
associated with platform discovery
...
e
...

0 ...


This type of model still requires trial error and it only learns when it is at the goal
...

Performance of simulations of the model
It learns stereotype trajectories rather
than optimal trajectories
...

Unfortunately the model doesn’t take
into straight
...

Place constant means the goal is
always in the same place, place
random means the goal gets moved
...

Evaluation
Captures the idea of ‘response’ (body turn) learning in basal ganglia, versus ‘place’ presentation in
hippocampus
...

The Brown and Sharp model performs complex ‘stimulus-response’ learning- takes into account place and
direction and then make a body turn of either left or right in order to get to the goal
...

Solves the temporal credit assignment problem by using the recency-weighted cumulative Hebbian value
associated with each connection
...
There may be something in the

15th November
synapses that governs LTP which takes into account decay
...
But then it is impossible for
one synapse to know all the other synaptic activity of the other synapses
...




---

Title: Hippocampal and striatal roles in spatial navigation
Description: Computational Science module for Neuroscience BSc at UCL Lecture by Prof Neil Burgess I got 69 in the module and a first class degree in Neuroscience

Buy These NotesPreview