MALENA (Machine Learning for Network Analytics)

Joint Research Team between Inria and IIT Bombay

• Presentation
• Team members
• Publications
• Visits
• Job offers
• Miscellaneous

MALENA (Machine Learning for Network Analytics) is a joint Indo-French research team between IIT Bombay and Inria. It was created in March 2017 and is financially supported for 3 years by DST, CEFIPRA and Inria.

We broadly plan to address three subareas:

Network inference: Estimating network characteristics such as diameter, edge conductances, average occurrence of specified motifs, testing hypotheses regarding graph structure, rumour source detection, network tomography, furthering ongoing work in some of these as well as some new directions. Our approach will be a blend of modern statistical tools and signal processing tools, exploiting structural aspects such as sparsity and correlations.

Network algorithms: This includes distributed algorithms on networks such as gossip based, distributed learning and optimization, and also algorithms for networks such as ranking, approximate computation of centrality measures, clustering etc. In addition to developing further Markov Chain Monte Carlo and related algorithms, we also hope to leverage techniques of high dimensional inference such as compressive sensing, low rank approximations, etc.

Dynamic and strategic issues: This will cover graph dynamics, dynamics on the graphs and related optimisation and estimation issues, strategic behaviour on networks, e.g., resource allocation, network based games and belief propagation. Paradigms from statistical physics such as metastability and cavity methods will be principal tools used in this subarea.

Back to the top

French side:

Konstantin Avrachenkov (Team leader, Inria NEO Team);

Eitan Altman (Inria NEO Team);

Marc Lelarge (Inria DYOGENE Team and Safran).

Indian side:

Vivek S. Borkar (Team leader, IIT Bombay);

Nikhil Karamchandani (IIT Bombay);

Sharayu Moharir (IIT Bombay);

Neeraja Sahasrabudhe (IISER Mohali).

• K.E. Avrachenkov, V.S. Borkar and S. Pattathil, Controlling G-AIMD by index policy, Proceedings of IEEE CDC 2017, Melbourne, Australia.
• J. Hachem, N. Karamchandani, S. Moharir and S. Diggavi, Caching with partial matching under Zipf demands, Proceedings of ITW 2017.
• A. Kalvit, V.S. Borkar and N. Karamchandani, MCMC approaches to rumor source inference using pairwise information, Proceedings of VALUETOOLS 2017, Venice, Italy.
• V.S. Borkar and K. Chadha, A reinforcement learning algorithm for restless bandits, Proceedings of Indian Control Conference, Jan 4-6, 2018, IIT Kanpur.

Back to the top

Visits from France to India:

• K.E. Avrachenkov visited IIT Bombay and TFIR from Oct 30 to Nov 4, 2017;

Visits from India to France:

• Akhil Thalasseryveettil (IIS Bangalore) internship from Feb 1 to July 31, 2017;
• Sarath Pattathil (IIT Bombay) internship from May 5 to July 5, 2017;
• Ranbir Singh (IIT Bombay) internship from Sept 11 to Oct 27.

We have PhD and PostDoc offers on the topic of the project. Please inquire at k.avrachenkov@sophia.inria.fr .

Back to the top

Back to the top