Recursive Markov chains in DNA->protein analysis, and particle physics. I am using the source for a protein sequence Hidden Markov detection program to make a program that modifies itself to model the Markov model and other techniques for its ability to correlate measured results to measured data. In the case of particle tracks it is a fairly simple P(A|B) since the freedom of possible precursors is low. DNA interaction is far more complex as it has a higher number of factorial interactions and thus can have patterns that can originate from two different models, but as time goes by it becomes more certain. I am also modifying GNUchess as a project in game theory. I hope to learn some new relationships as I create a Markov player that uses stochastic tools as well as others to dominate the GNUchess automaton. I can beat the computer chess player, but perhaps only because of limited depth. I am making a learner that works generally with proteins, particles and games and hoping this is general enough to have a global minimum method that is universally applicable.

The goal is to create a Markov process that measures itself and works from improbability as well as probability. By approaching from two directions it would seem that a probability vector space could be established outside the framework to establish a generalized to specific search focus that goes from known to unknown condition as a certain local minimum without a complete solved brute force model. One of the original sources that is well structured comes from HMMER

In a strange way, XML files of SVG images are a Hidden Markov Model of the final image presentation. I can see that compression of information is higher when the appropriate relevant Markov chains are considered. Temporally and practically it is not even advantageous to have more information than is absorbed. It would be comparable to creating a desired molecular transition that is sequential and any additional energy or information simply serves to degrade the reliability of the conclusion. Information could be considered to be energy, but like energy, information is relative and as a result is not absolute in analysis. Since mv is relative, then 1/2mv^2 is relative and thus transition differential with respect to time is relative. So information or relative change is not an absolute measurable quantity as it is observer dependent and thus has infinite possible solutions. To analyze the information content it must be done with respect to a frame of reference. Even very simple relationships can have a varying degree of recovered information based on the observer.

## 0 comments:

Post a Comment