Debarghya Ghoshdastidar

I joined Theory of Machine Learning group in May 2016 as a post-doctoral researcher. I am also a junior fellow in Research Unit 1735: Structural Inference in Statistics: Adaptation and Efficiency under the German Research Foundation. Broadly, I am interested in (i) developing learning algorithms for the purpose of network analysis, and (ii) understanding and applying results from random matrix theory and random graph theory. My current research revolves around studies related to hypothesis testing and clustering in the context of random graphs and hypergraphs.

Background

• Ph.D. in Computer Science & Automation from Indian Institute of Science, Bangalore, India
  Thesis: Consistency of spectral algorithms for hypergraphs under planted partition model pdf slides
• Masters (Engineering) in System Science & Automation from Indian Institute of Science, Bangalore, India
  Thesis: Properties of multivariate q-Gaussian distribution and its application to smoothed functional algorithms for stochastic optimization pdf slides
• Bachelors (Engineering) in Electrical Engineering from Jadavpur University, Kolkata, India

Publications

• D. Ghoshdastidar and A. Dukkipati. Uniform hypergraph partitioning: Provable tensor methods and sampling techniques. Journal of Machine Learning Research 18(50), pp. 1−41, 2017. link arXiv code
• D. Ghoshdastidar, M. Gutzeit, A. Carpentier and U. von Luxburg. Two-sample tests for large random graphs using network statistics. COLT, 2017. link arXiv slides
• S. Haghiri, D. Ghoshdastidar and U. von Luxburg. Comparison based nearest neighbor search. AISTATS, 2017. link arXiv
• D. Ghoshdastidar and A. Dukkipati. Consistency of spectral hypergraph partitioning under planted partition model. The Annals of Statistics, 45(1), pp. 289-315, 2017. DOI arxiv
• D. Ghoshdastidar, A. P. Adsul and A. Dukkipati. Learning with Jensen-Tsallis kernels. IEEE Transactions on Neural Networks and Learning Systems, 27 (10), pp. 2108-2119, 2016. DOI pdf code
• A. Dukkipati, D. Ghoshdastidar, and J. Krishnan. Mixture modelling with compact support distributions for unsupervised learning. IJCNN, 2016. DOI
• D. Ghoshdastidar and A. Dukkipati. A Provable Generalized Tensor Spectral Method for Uniform Hypergraph Partitioning. ICML, 2015. link slides
• D. Ghoshdastidar and A. Dukkipati. Spectral clustering using multilinear SVD: Analysis, approximations and applications. AAAI, 2015. link slides
• D. Ghoshdastidar and A. Dukkipati. Consistency of spectral partitioning of uniform hypergraphs under planted partition model. NIPS, 2014. link poster code
• D. Ghoshdastidar, A. Dukkipati, A. P. Adsul, and A. S. Vijayan. Spectral clustering with Jensen-type kernels and their multi-point extensions. CVPR, 2014. DOI arxiv
• D. Ghoshdastidar, A. Dukkipati, and S. Bhatnagar. Newton based stochastic optimization using q-Gaussian smoothed functional algorithms. Automatica, 50 (10), pp. 2606-2614, 2014. DOI arxiv code
• D. Ghoshdastidar, A. Dukkipati, and S. Bhatnagar. Smoothed functional algorithms for stochastic optimization using q-Gaussian distributions. ACM Transactions on Modeling and Computer Simulation, 24 (3), Article 17, 2014. DOI arxiv code
• A. Dukkipati, G. Pandey, D. Ghoshdastidar, P. Koley, and D. M. V. Satya Sriram. Generative Maximum Entropy Learning for Multiclass Classification. ICDM, 2013. DOI arxiv slides
• D. Ghoshdastidar and A. Dukkipati. On power law kernels, corresponding Reproducing Kernel Hilbert Space and applications. AAAI, 2013. link slides
• D. Ghoshdastidar, A. Dukkipati, and S. Bhatnagar. q-Gaussian based Smoothed Functional algorithms for stochastic optimization. ISIT, 2012. DOI arxiv

Preprints / Workshop papers

• D. Ghoshdastidar, M. Gutzeit, A. Carpentier and U. von Luxburg. Two-sample hypothesis testing for inhomogeneous random graphs. arXiv preprint: 1707.00833v2. arXiv
• D. Ghoshdastidar and A. Dukkipati. Coloring random non-uniform bipartite hypergraphs. arXiv preprint: 1507.00763v2. arXiv
• D. Ghoshdastidar and U. von Luxburg. Do nonparametric two-sample tests work for small sample size? A study on random graphs. NIPS-2016 workshop on Adaptive and Scalable Nonparametric Methods in ML.

Invited talks

• September 2016: Dagstuhl seminar on Foundations of Unsupervised Learning
  Title: Two sample tests for large random graphs.
• January 2016: Department of Computer Science, IIT Bombay
  Title: Consistency of spectral hypergraph partitioning under planted partition model.
• October 2015: Department of Computer Science, R. V. College of Engineering, Bangalore
  Title: Recent trends in AI.
• March 2015: IKDD Conference on Data Sciences (CoDS 2015)
  Title: Consistency of spectral partitioning of uniform hypergraphs under planted partition model.
• October 2012: Bangalore Probability Seminar
  Title: On some statistical properties of multivariate q-Gaussian distribution and its application to smoothed functional algorithms.