Multiverso is a parameter server based framework for training machine learning models on big data with numbers of machines. It is currently a standard C++ library and provides a series of friendly programming interfaces. With such easy-to-use APIs, machine learning researchers and practitioners do not need to worry about the system routine issues such as distributed model storage and operation, inter-process and inter-thread communication, multi-threading management, and so on. Instead, they are able to focus on the core machine learning logics: data, model, and training.

For more details, please view our website http://www.dmtk.io.

Linux (Tested on Ubuntu 12.04)

1. Run cd third_party; ./install.sh to install the dependence.
2. Run make all -j4 to build the multiverso.

Windows

For windows users, please refer to README in windows folder.

Current distributed systems based on multiverso:

• lightlda: Scalable, fast, lightweight system for large scale topic modeling
• distributed_word_embedding Distributed system for word embedding
• distributed_skipgram_mixture Distributed skipgram mixture for multi-sense word embedding