Thalamo-Cortical Modeling and Implementation workgroup 2011

Important links:

• Requirements (Hardware and software info)
• Details of the model will be put up soon.

Ideas for projects

Below are a list of stimuli that can be put to test with the network at its current state:

• simple sweeps // or \\.
• < , > like stimuli.
• periodic sinusoid or /\ (sawtooth like)
• More realistic stimuli ... (like cochlear spikes with low Q)

Open questions :

• A good/simple readout.
• How does the model deal with stimuli with slow/no variations for longer periods.
• How about learning more than one feature ?? How can competition be put to learn the whole set of basis functions ?

Out of the box :

• Since we are essentially learning temporal sequences, can the same modeling basics be used to do visual gesture recognition ?
• Is it really necessary to have an orderly delays ? Can it just be any random delays (or any random order of frequency channels). There has already been some preliminary investigation in this direction since the original 'big' model has a range of delays.