Читать книгу Informatics and Machine Learning. From Martingales to Metaheuristics онлайн
96 страница из 101
3.4 Sequential Processes and Markov Models
Just as ssss1 finished with a Math review, we do the same again here in the context of sequential processes. The core mathematical tool for describing a sequential process (where limited memory suffices) is the Markov chain, so that will be defined first. In the context of genome analysis, however, the standard Markov chain based feature extraction is no longer optimal (especially given the nature of the computational resources). Thus, novel mathematical generalization of the Markov chain description, interpolated Markov models, will be given as well.The gap/hash interpolated Markov model, in particular, can be used to “vacuum‐up” all motif information in specified regions. This could be used and directly integrated into an HMM‐based gene finder (ssss1 and ssss1), or, alternatively, provide identification of a typical motif set for some circumstance (as will be done in ssss1).
3.4.1 Markov Chains
1231234n
11
ii − 11ii − 1i − 1, ii − 1, iL,L − 111i = 2…Li − 1, iyxyyxyi − 1, ixyxyyyxyyxyyi − 1, ixyyxyxxyxy