 |
Center for Nonlinear Studies |
Efforts to model recognition processing in the brain have reached and important milestone. The most advanced computers can now meet or exceed the computational scale of the brain. However, they still do not match the function. Some of the problems that were posed in the field's early days still haunt models today. In essence these are scalability problems where the number of required resources (e.g. theoretical neurons) increases exponentially with the number of objects to be recognized; to the point that such models are not plausible for realistic scenarios. I will give an overview of these problems, strategies to overcome these limitations, and what I think makes it difficult to move past these problems
Host: Peter Loxley, loxley@lanl.gov