— Research
Three research projects and two published papers, in Gaussian processes, optimal transport and energy systems. Each project has its own reader with the paper, a timeline, the reasoning behind it and its pipeline.
01 Research projects
Replicates Bachoc et al. (2018) on GP regression with probability distributions as inputs, then extends it to Wasserstein Tangent Gaussian Processes.
Open reader →Treats the joint price–production forecast as a probability measure so that the cannibalisation effect in renewable markets enters the GP.
Open reader →Corrects the geometry of diffusion-based video summarisation: barycenter consensus, Dirichlet noise and a Wasserstein loss.
Open reader →02 Publications
Milind Sahu, Darun, Latheesh Roy
A GP-based interpolation scheme for implied volatility surfaces that embeds no-arbitrage constraints directly in the covariance structure, avoiding post-hoc smoothing or projection.
DOI 10.13140/RG.2.2.33170.62400 ↗Milind Sahu, Darun A. · Axiom Lab · CC BY-NC 4.0
Introduces the Civic Energy Disaggregation Mandate (CEDM), arguing for NILM as public utility infrastructure rather than opt-in consumer technology, with a five-model ML pipeline benchmarking disaggregation accuracy and deployment feasibility.
DOI 10.13140/RG.2.2.20940.17283 ↗03 Foundations