Research papers and preprints I have contributed to, spanning machine learning, optimisation and applied probability.
Rushil Gupta, Sandeep Juneja
A path-splitting method for estimating terminal distributions of stochastic differential equations under a fixed simulation budget. We characterise asymptotic errors and develop a splitting strategy that reduces mean Kolmogorov–Smirnov error by 10–25% in many settings compared with independent sampling.
Saptarishi Dhanuka, Sarvesh Iyer, Manmeet Singh, Mihir More, Rushil Gupta, Dhruman Gupta, Parthasarathi Mukhopadhyay, Sandeep Juneja
An adaptive framework that mixes many probabilistic weather forecasts using both machine learning and a mixture of experts. We prove logarithmic regret against the best static mixture of experts in hindsight, and show empirical gains on temperature forecasting.
Sarvesh Ravichandran Iyer, Himadri Mandal, Dhruman Gupta, Rushil Gupta, Agniv Bandhyopadhyay, Achal Bassamboo, Varun Gupta, Sandeep Juneja
A sharp characterisation of how accurately a self-normalized importance sampler can approximate a tilted distribution from samples of the base distribution alone. Bounded random vectors need polynomially many samples in the tilt amount; unbounded ones need super-polynomially many.