Software

Smart broadcasting

Check out the implementation of two smart broadcasting algorithm, which are able to optimize when to post to achieve higher visibility in a social network

You can find more information about fair classifiers in our KDD 2016 and WSDM 2017 papers. Please, feel free to send any suggestions, comments, bugs or alternative implementation to utkarshu[at]mpi-sws.org

Computational discrimination

Check out the implementation of fair logistic regression, which is able to provide predictions that do not discriminate with respect to one of more sensitive attributes

You can find more information about fair classifiers in our AISTATS 2017 and WWW 2017 papers. Please, feel free to send any suggestions, comments, bugs or alternative implementation to mzafar[at]mpi-sws.org

Recurrent Marked temporal point processes

Check out the implementation of the recurrent marked temporal point processes, which are able to accurately predict times and marks of general marked temporal point processes

You can find more information about recurrent temporal point processes in our KDD 2016 paper. Please, feel free to send any suggestions, comments, bugs or altternative implementations to hanjundai[at]gatech.edu

Information diffusion and network evolution

Check out the implementation of COEVOLVE, our joint point process model of information diffusion and network evolution.

You can find more information about the framework in our NIPS 2015 paper. Please, feel free to send any suggestions, comments, bugs or alternative implementation (Python, etc.) to mehrdad[at]gatech.edu.

Activity shaping

Check out the implementation of our social activity shaping convex optimization framework. It uses Hawkes processes to model endogenous and exogeneous activity in a social network.

You can find more information about the framework in our NIPS 2014 paper. Please, feel free to send any suggestions, comments, bugs or alternative implementation (Python, etc.) to mehrdad[at]gatech.edu.

Influence estimation

Check out the implementation of ConTinEst. It is a highly efficient influence estimation algorithm for continuous time diffusion networks. ConTinEst uses randomization to scale influence estimation to networks with million of nodes and it can be used as a building block of InfluMax.

You can find more information about the algorithm in our NIPS 2013 paper. Please, feel free to send any suggestions, comments, bugs or alternative implementation (Python, etc.) of ConTinEst to dunan[at]gatech.edu.

Network Inference

Check out the implementations of NetInf, NetRate, MultiTree and InfoPath. The four of them are network inference algorithms from diffusion traces. NetInf and MultiTree exploit submodularity and NetRate and InfoPath exploits convexity to make the inference problem tractable. NetInf, MultiTree, and NetRate support static networks, while InfoPath supports both static and dynamic networks

You can find more information about the algorithms in Publications. Please, feel free to send any suggestions, comments, bugs or alternative implementation (Python, etc.) of NetInf, NetRate and/or MultiTree to manuelgr[at]mpi-sws.org.

Influence maximization

Check out the implementation of InfluMax. It is a influence maximization algorithm for continuous time diffusion networks. InfluMax exploits submodularity and is capable of evaluating influence using CTMCs.

You can find more information about the algorithm in our ICML 2012 paper. Please, feel free to send any suggestions, comments, bugs or alternative implementation (Python, etc.) of InfluMax to manuelgr[at]mpi-sws.org.