Discussion/Questions After the Talk.
Dana: If surfaces are creating the percept of an image space, it seems what you are calling generic views are quite in line with Bayes.
A: When I talk about Bayes, the priors are stochastic and non ergodic. Also the Kanizsa diagrams show that we are not using the priors. Bayesian formulation is far too unconstrained and general.
Comment(Jeff?): I see most of the figures as ambiguous.
Q: Isn't the generic view principle a prior?
(Comment): Bill Freeman has a paper which said the generic view principle in terms of Bayes.
Q: Most of the things can be explained because you see motor experience and not just image ?
Comment: We see the impossible triangle through bottom up process even though we know at the top that its not possible.
Q (Jeff): You are confusing with models and theory. Do you think there is a neurobiological explanation for all this.
Answer : Reductionism is metaphysical.
Comment(Dana): Through development we get the the kinetic depth and stereo vision. We might not be able to get this figured out until we track the developmental trajectory. Also perception and motor systems do interact.
Comment(Jeff): All the speakers yesterday were in the optimistic camp and today we have the speakers in the pessimistic camp
Back to the
main index
for Inference and Prediction in Neocortical Circuits.