mrjiaruiwang
commented
9 years ago This is an interesting point. The mathematics do not care about what network it is being used for, but the person using the mathematics should care about what mathematics he or she is using for what purpose. It is my personal opinion that SVMs tend to be overused because it is usually the first thing people go to for trying to infer patterns from data. However, there is a discrepancy between the "hidden" model that SVMs try to approximate versus the intuition we have about biology. Rarely in biological circumstances do we see linear combination or a particular kernel function as a model for input/output. I feel that Graphical models fit the brain much better than SVMs because at least we have the biological basis to say that neurons are connected in a fashion similar to other networks versus say genomics where it is not yet clear how genes are connected. The example we went over in class with the "mixture" of gaussians was interesting, but I wonder if there is biological justification for such a framework.
-We've talked about how the math doesn't care about what network we are looking at, whether it be social or brain or something else. How do interpretations of graphs change based on what type of network you look at?