describe the equation principle ?
- Start: a1 -> w1 -> transfer function -> activation function -> output function
What is meant by the small a in the equation principle ?
Small a = Input (activity)
What is meant by the lare w in the equation principle ?
- large W = weight
What is meant by the transfer function in the equation principle ?
- All the inputs multiplied by the weights and then add each other up
- a1xw1+a1xw2…
What is meant by the activation function in the equation principle ?
- Different functions representing the activation function
- linear, threshold binary and sigmoid
What additional unit is presented in the equation principle ?
- The bias
What is the formula of the hebbian learning rule ?
wij + Deltawij = new wij
- wij = current weight (given)
- Delta wij = the new wight change
How do we calculate the delta wij ?
- activity of neuron I multiple by activity of neuron j multiple by a learning constant
- everything is given
Out of how many cells does the Hippocmapus exist ?
- 2:
- > Cornu Ammonis (CA)
- > Dentate gyrus
What is the main function of the Hippocampus ?
- It is mainly active during episodic memory
- memory which u can activly recall
- mainly involved i formation of memory
How is episodic memory formed ?
- The HC builds a formation
- The formation is needed to create episodic memory
- it needs a rapid combination of information from dif sources -> This means formation
How is procedural memory formed ?
- gradual development over many related experiences
- skill learning
How does the information flow outside of the HC ?
- Input zone for HC: is the Parahippocampal gyrus and entorhinal cortex
- output zone: back projection to cortical areas where the input came from
What is meant by the enthorinal cortex ?
- Cortex/network for memory, navigation-> input for HC
What is in general the internal path of the HC
- Information processing in 3 stages: starts at the perfront path -> dentate gyrus -> CA3 -> CA1
Explain the general internal path of the HC a bit more in detail:
- Perfronth path synapse with cells in dentate gyrus
- output from dentate gyrus via mossy fibers toward cell in CA 3
- some cells might be directly connected from perforant path to CA3 cells
- Recurrent connections: cells in CA 3 send infomation back and forward
- Schaeffer collateral cells -> send the outpuf from CA3 to the CA 1
- CA 1 send its input to the neocortex/subiculum structure
- Perfronth path synapse with cells in dentate gyrus
What is so special between the connection btw the dentate gyrus and CA 3 cells ?
- Different input patterns activate different CA3 cells
- Only some cells are active at a certain time (called sparse)
- The sparse is need to activate the auto associator which makes it possible to store more memory
How does the hippocampus learn ?
- Also via the hebbian learning rule
- it works really quickly less then a second which is needed for episodic memory
Where does compettive learning occur ?
- In DG and CA1
- the output from the DG system will be less correlated and more categorised than the inputs to it from the perforant path
How does the hippocampus keeps out irrelevant information ?
- Via threshold which is connected to Information about rewards and punishments by the amygdala
- is in the enthorinal cortex 5%
Give a short summary of what is going on in each stage of the hippocmapus:
- competitive learning in DG,
- autoassociation in CA3,
- competitive learning in C A l
- pattern association between CAl and entorhinal cortex.
- competitive learning in DG,
What happens if we bases a neuronal network on the HC ?
- it can stores a large number of unrelated patterns with a single presentation
- and it can retrieve the stored patterns from fragments of them
- so it works just like the HC
What is the sum of external and internal input in a autoassociator cell ?
- netinput
Learn something about the sparse input from the denatte gyrus:
- sparse input is allways 5% for CA1 and DG
- Only in CA 1 there it is 1%
- sparse input is the reason why we can store so many information (SPARSNESS)
