Computational Challenges (and limitations) for the Cortex

This lecture by Leslie Valiant on 'What are the Computational Challenges for the Cortex' is a followup lecture to yesterdays talk by Edvard Moser.  It focuses on the theoretical limitations of cortex models based on known physics and biological constraints in neural assemblies.

Over a lifetime the brain performs hundreds of thousands of individual cognitive acts of a variety of kinds, including the formation of new associations and other kinds of learning. Each such act depends on past experience, and, in turn, can have long lasting effects on future behavior. It is difficult to reconcile such large scale capabilities quantitatively with the known resource constraints on cortex, such as low connectivity. Here we shall describe an approach to this problem, in terms of concrete functions, representations, and algorithms, that seeks to explain these phenomena in terms that are faithful to the basic quantitative resources available. Until recently an algorithmic understanding of cognition has been regarded as an overambitious goal for experimental neuroscience. As we shall explain, with current experimental techniques this view is no longer justified, and we should expect algorithmic theories to be experimentally testable, and tested.

This lecture is part of the Heidelberg Laureate Forum, so all of the people involved are Nobel prize winners or Turing award winners or the equivalent, so everything is really top notch.  We'll be following up later this week with another talk by Yoshua Bengio from the same forum.



Observations

1:  I have to be honest, at the end of this talk i was kind of disappointed, because he sets up this whole scenario where he shows the math leads to certain conclusions or constraints, and then you are basically left hanging with no conclusions at all.  I wanted another 10 minutes where he would actually lay out an actual theoretical model we could explore and test.

However, something he pointed out in the talk got me thinking about a really interesting new approach to building neural net models.  It ties into some work i did on fast multi-dimensional scaling algorithms in the late 90s.  If you run through the section of the lecture that talks about sparse random connections you might start to grok it.

If my very preliminary thinking on this pans out at all, i'll do a HTC post where i describe the algorithm in more detail.

2:  Note that sparse connections are a key component of what the thousand brains theory is proposing.

Comments

Post a Comment

Popular posts from this blog

CycleGAN: a GAN architecture for learning unpaired image to image transformations

Image
 Like the Doublemint Twins touting the joys of Doublemint gum, 2 GANs are surely better than 1 GAN.  Especially if we package them together inside of one meta GAN module.  And this is exactly what the CycleGAN architecture does. Have you ever harbored dark secrets of turning a horse into a zebra?  The CycleGAN was developed to do just that. Learn how to turn a horse into a zebra.  And more. Now right away you can notice a difference between the image to image transformation GAN architectures we've been discussing over the last few posts.  Those last few posts described systems that learn from a database of matched input-output image pairs. And if your goal is to turn an edge representation into a nicely filled in continuous tone image, it's easy to build your database of matched input-output image pairs that your GAN system can then learn off of.  Take a continuous tone image (which will be the output of the database pair entry), then run it through an edge detector algorithm. 
Read more

Pix2Pix: a GAN architecture for image to image transformation

Image
 I thought following up yesterday's TraVelGAN post with a Pix2Pix GAN post would be useful to compare what is going on in the 2 architectures.  Two different approaches to the same problem. I stole this Pix2Pix Overview slide below from an excellent deeplearning.ai GAN course (note that they borrowed it from the original paper) because it gives you a good feel for what is going on inside of the Pix2Pix architecture.    Note how the Generator part is very much like an auto-encoder architecture, but rebuilt using the U-Net architecture features (based on skip-connections) that fastai has been discussing in their courses for several years before it became more widely known to the deep learning community at large  (and which originally came from an obscure medical image segmentation paper) . So the Generator in this Pix2Pix GAN is really pretty sophisticated, consisting of a whole image to image auto-encoder network with U-Net skip connections to generate better image quality at highe
Read more

Smart Fabrics

Image
"What makes smart fabrics revolutionary is that they have the ability to do many things that traditional fabrics cannot, including communicate, transform, conduct energy and even grow."  This is a quote from Rebeccah Pailes-Friedman, who is a professor at Pratt with a research focus on developing wearable technology and smart textiles. Enhancing textiles can be viewed through different lens.  Enhancements can be aesthetic, or performance boosting, or durability boosting. 
Read more