Yes, great presentation and summary of the work done by these guys at CMU. It reinforces my belief that sensory processing should be explained in terms of finding efficient representations, in other words: dimensionality reduction. One nice aspect is that there are machine learning techniques that do exactly this in a general way, like(Restricted) Boltzmann Machines or adapted ICA. The wonderful thing about this kind of architecture is that at the same time it explains biological phenomena, and performs very well on computer vision tasks like classification and clustering. I'm wondering how much further one could extend such architecture to coding of spatiotemporal (video) patterns, multimodal patterns (video + audio) and eventually coding of 3D objects. They are all 'just' extensions of such a model, 'just' about finding efficient ways of learning the joint probability distributions :) however I imagine that finding efficient ways of training such models (e.g. finding compact representations) should become increasingly hard.
The 'ultimate' general computer vision architecture, in my mind, would be series of (stereo) images as input, learns such a hierarchical model (using dimensionality reduction) and uses this to interpret the images, and in parallel uses this 2D/2.5D interpretation (plus disparity information) to create a full dynamical 3D mapping of its environment, which in term could be subject to further interpretation (dimensionality reduction). A huge task :) But, increasingly realistic. This, then, of course, could be used by e.g. Novamente to step out of the virtual realm and perform inference in the real world. And that's where the fun would start, don't you all agree? On Feb 17, 2008 3:16 PM, Bob Mottram <[EMAIL PROTECTED]> wrote: > On 17/02/2008, Vladimir Nesov <[EMAIL PROTECTED]> wrote: > > This is a tutorial talk I saw recently (see Section A): > > http://nips.cc/Conferences/2007/Program/event.php?ID=577 > > Michael Lewicki > > Sensory Coding and Hierarchical Representations > > I like this guy's approach, which covers many familiar topics. > Particularly I liked the way he treated both vision and auditory > systems as essentially the same problem. > > > ------------------------------------------- > agi > Archives: http://www.listbox.com/member/archive/303/=now > RSS Feed: http://www.listbox.com/member/archive/rss/303/ > Modify Your Subscription: http://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com > ------------------------------------------- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
