Notes from Breakout Session II

The four groups met separately to prepare responses to the comments from the previous day.

Reply From Psychologists

The theorists wanted to be hired. The primate experiments would go much better if they did psychophysical experiments.

Giving away raw data looks like giving away ones own work.

Alessandra: Once something is published it is public domain and so the data should me bade public.

Judith: It could be that you have one sort of analysis and more analysis might be waiting. you just don't want to give away the data.

We should be able to publish the data

From Theory group I.

• Are there associative spiking models with spiking neurons? (Yes). Fritz will give a presentation on that.
• Theoreticians will become aware of training in experimentation. This will give us more intuition. But we need more help in doing that. We would also like the data from you!
• Russ: Rigorous model comparisons between different models. One can apply the same inputs to models and get the results. Should document when the models fail.
• Falsifiable predictions? : We alread do.

• The job of theorists is not just to make testable hypothesis.

From Anatomists

• Now more labs are producing less descriptive and more quantitative data. There is some kind of questions that take too much time to address. (For eg: number of synapses). How can computer science help us for this?

• There are computer vision techniques and tools available that could be used.
  Multi-electrode recording? Its a lot of effort. If the cells cannot be activated at the same time, it amounts to recording one after another.

• Natural stimuli.
  We have acquired multichannel recording systems and we are racing to do multi electrode recording.

From Theory II

• Bayesian Framework. : Bayesian framework is a principled way of looking at the computational problems faced by the brian given the ambiguous

  Priors can be implemented as lateral connectivity or synaptic weights or top-down predictions.

  Predictions for testing the bayesian framework: Rajesh Rao's work

• Different forms of feedback : We really don't know how these forms of feedback are. The diffused feedback that is topographic could be doing an spatial priming and the patchy ones are doing feature priming.

• Tony: Natural stimulii. Adaptive models: Natural stimulii. Yes.

  In the case of hardwired models: it will give tighter prediction to the experimentalists.

Back to the main index for Inference and Prediction in Neocortical Circuits.