Home Page of Subhro Ghosh


Subhroshekhar Ghosh
National University of Singapore
Department of Mathematics
Dept of Statistics and Data Science
Institute of Data Science
Singapore

Email:
subhrowork (**at**) gmail.com
matghos (**at**) nus.edu.sg




Research   |   Selected & recent works   |   Publications   |   Teaching   |   Students & postdocs   |   Grants & organisation   |   Seminar   |   Gallery



About me


I am an assistant professor at the National University of Singapore and a faculty affiliate at the Institute of Data Science, NUS. I am broadly interested in stochastics, focussing on problems from statistical physics and the math of data, and their interactions. Before joining NUS, I was a post doc at Princeton University, and prior to that I obtained my PhD from the University of California, Berkeley. Earlier, I received my Bachelor in Statistics and Master in Mathematics degrees from the Indian Statistical Institute.



Research


I am broadly interested in stochastics, focussing on problems from statistical physics and the math of data, and their interactions. These encompass constrained stochastic systems and their applications, including problems of learning under complex structure (e.g., latent group actions), dimension reduction, sampling and optimization, statistical networks and signal processing. Key paradigms include determinantal processes (DPP), strong Rayleigh measures and negative dependence, multi reference alignment (MRA), empirical likelihood, generative priors, Gaussian analytic functions (GAF), random tensors and stochastic geometry. The investigation of these problems naturally brings together a wide array of tools and techniques, including probability, harmonic and complex analysis, persistent homology and the theory of group representations.

For more on my research, please refer to
Selected & recent works and my full list of Publications.



Some selected and recent works




Publications




A Random Gallery


            Left to right: Surface plot of normalised GAF; and Fractal Gaussian networks


            Left to right: Gaussian DPP for clustering Fisher's Iris data; and stochastic geometry of spectrogram level sets


            Left to right: Laplace transform based sampling for UK retail data; and DPP based minibatch sampling for Stochastic Gradient Descent


            Left to right: Conditional intensity for a Gaussian matrix and for a Gaussian polynomial


Back to top



Teaching


Some courses that I have taught:



Research Grants and organisational activities


My research work and organisational activities have been generously supported by :



Graduate students


I have had the pleasure to supervise :

Research Fellows


I have had the pleasure to mentor :

Seminar


I organise the Stochastics Seminar together with Ujan Gandopadhyay. If you are interested in giving a talk, please reach out via e-mail.

Back to top