The Idea of Combining Random Matrix and Graphical Models in Machine Learning

Random matrix is an astonishing development in math. It is very strong for description a stochastically system, especially for random dynamic system. We can even try to research the ultimate problem of a system; like, the light speed for universe, the acceleration of gravity for earth. How to descript a stock trading system as a physical system? Stock trading system is too complex; it will make most people like blind men touching an elephant when they try to answer the fundamental question like what is the stock trading system. This is like stupid problem. Is it really stupid? Then what is time in physics system? It will need the brain like Newton and Einstein. To answer the very fundamental problem of the complex stochastically dynamic system like stock trading system, maybe it is a way that we try to find invariant   or invariant   structure using the random matrix. I have got some results about it all by myself.

Graphical model is a amazing idea and very powerful for the machine learning and, which is the most important tools in 3rd generation development of machine learning. It is very good for statistical reasoning, but for application in complex dynamic system(like stock trading system), more tricky will be need, like HMM.

Another problem, most of time, we want to find some knowledge from a random system. How do we use the random matrix for reasoning and machine learning? I think apply random matrix in machine learning will induce more ideas from modern physics to machine learning, business intelligence, and data mining, and promote the research and application of machine learning to a new stage.