Virtual Statistical Mechanical Neural Computer

摘要

The study applies recently developed statistical algorithms and an innovative large system scaling technique to the problem of implementing a virtual neural computer. Once the techniques are shown to faithfully model highly nonlinear, nonequilibrium probability distributions, they are applied to the brain. The statistical mechanical neural computer (SMNC) developed in this thesis makes use of scaling to effectively filter the information flow and to model its contents. The implications for command and control are the SMNC's ability to recognize patterns of previously stored information detecting similarities between new and old data. The purpose of the SMNC is to serve as a decision aid that will contain high quality information about specific nonlinear relationships related to system variables, through the aggregation of information into coarse-grained data at a mesoscopic level. This should give the user, be it battlefield commander or Wall Street analyst, the ability to more accurately forecast the most likely course of events a given scenario would follow based on its recent history. The SMNC is well suited for the study of stochastic processes. Its methods for the aggregation and scaling of data make it an effective tool for the study of both short and long term behavior in nonlinear nonequilibrium systems.