Anant Mathur
Researcher in Statistics, Machine Learning & Optimization
I'm a researcher working in statistics, machine learning and optimization. I completed my PhD at the University of New South Wales, Sydney, supervised by Zdravko Botev and Sarat Moka, where my research centred on sparse methods for high-dimensional data.
Research interests
Statistical modelling
Statistical methods for matrix, regression, and related learning problems. My PhD focused on structured variable selection: best subsets in generalised linear models, column subset selection, low-rank matrix approximation, and variance component models.
Model compression
Sparse optimisation. In my PhD I developed scalable optimisation methods for sparse, structured models, including:
- Majorisation–minimisation algorithms for convex ℓ₁-constrained problems
- Binary relaxation methods for non-convex ℓ₀-constrained problems
Efficient deep learning. I'm also interested in applying these sparse optimisation methods to neural network pruning, in particular constrained pruning under FLOP and latency budgets, to produce smaller, faster networks without sacrificing accuracy.
Applied interests
Methods from statistics and applied mathematics more broadly, for large-scale genomic data (particularly genome-wide association studies), quantitative risk management, and transportation optimisation.
Software
- combss — best subset selection in generalised linear models via continuous optimisation.
Latest news
Education
-
PhD (Statistics) — UNSW SydneySep 2020 – Jun 2025
-
BSc (Hons), Statistics — UNSW Sydney2015 – 2019