Markov Chains Latent Representation by Micro-similarities

A C++ implementation of various and new methods of representing Markov Chains in a latent space that hypothetically relates the Markov model in hand with multiple pre-trained Markov models. An example usage for this representation is to base the classification of sequences using the latent representation then applying a secondary classifier, as an alternative to classifying sequences using the maximum propensity method. The current results show that the prediction performance is improved while retaining the explanatory power of the generative models.

Installation on Linux

Prerequisites

• Install Python-64bits or run apt-get install python3 python3-pip, 32-bits version won't work. Python is used for Conan package manager installation.
• Install conan, c++ package manager, preferably running pip install conan. For more information and alternative installation options, please refer to conan manual page.
• Install CMake for generating the build script.

Building

• Since the project in experimental development and big changes may take place between successive versions, make sure you are on the desired version (tag). For example, git checkout embc. Moreover, the table in Applications lists various tutorials and each is associated with a particular version (tag).
• Run conan remote add bincrafters https://api.bintray.com/conan/bincrafters/public-conan
• Run conan remote add a-alaa https://api.bintray.com/conan/a-alaa/public-conan.
• In this repository folder create build folder and, after moving into the new folder, run conan install .. --build missing -s build_type=Release -s compiler.libcxx=libstdc++11.
• Run cmake .. -DCMAKE_BUILD_TYPE=Release.
• Run make -j8.

Installation

• Run make install from the build directory. By default, this will require root privileges, unless you change the CMAKE_INSTALL_PREFIX variable to a local directory.

Applications

The following tutorials (will be updated regularly), demonstrates the various applications of this project:

Tutorial name	Supported versions (tags)
Tutorial on Benchmarking Performance of Protein Subcellular Localization Prediction by Different Methods	embc

LICENSE

This project uses the MIT license. Please see LICENSE.txt in the main directory for more details.