The chmmpp software supports the analysis of multivariate time series data to detect patterns using a Hidden Markov Model (HMM).
Many applications involve the detection and characterization of hidden or latent states in a complex system, using observable states and variables. The chmmpp software supports inference of latent states integrating both (1) a HMM and (2) application-specific constraints that reflect known relationships amongst hidden states. For example, HMMs have been widely used in natural language processing to tag the part of speech of words in a sentence (e.g. noun, verb, adjective, etc.). But in many applications there are known relationships that need to be enforced, such as the fact that a simple English sentence must contain at least one noun and exactly one verb.
The chmmpp software supports both application-specific and generic methods for constrained inference. This includes a framework for customized Viterbi methods, constrained inference of hidden states with A-star and integer programming methods, and various contraint-informed methods for learning HMM model parameters. A focus of chmmpp is support for generic methods that enable the agile expression of complex sets of constraints that naturally arise in many real-world applications. Optimization constraints can be expressed in chmmpp directly in C++ or using the coek modeling framework. A variety of commercial and open source source optimization solvers can be used to ensure that maximum likelihood solutions are found for hidden states.
To install the library create a directory called build, navigate to this directory, and run cmake ..
A libary will be created in build/library and executables of the examples will be found in build/examples