Akshat Valse

Stochastic foundations of machine learning · Iowa State University

CV

Curriculum vitae

Stochastic analysis for machine learning: Langevin sampling and its discretization, large deviations and rare-event simulation, and the stability of learned PDE solvers.

Education

Education

Iowa State University, Honors Program

B.S. Statistics · B.S. Mathematics · B.A. Computer Science · B.S. Data Science

In progress, Fall 2026 · all five at the PhD level (6000-series)

  • Foundations of Probability Theory, measure-theoretic STAT 6541
  • High-Dimensional Probability and Linear Algebra for Machine Learning MATH 6230
  • Advanced Topics in Machine Learning COM S 6730
  • Advanced Topics in Artificial Intelligence COM S 6720
  • Advanced Topics in High-Performance Computing COM S 6250

Planned, Spring 2027

  • Advanced Probability Theory
  • Advanced Stochastic Processes

Graduate and advanced coursework, completed

Probability and statistics
  • Theory of Probability and Statistics STAT 5430
  • Stochastic Process Models STAT 5540
  • Empirical Methods for the Computational Sciences STAT 5830
  • Multivariate Analysis STAT 5750
  • Categorical Data Analysis STAT 5770
  • Statistical Computing STAT 5860
Mathematics
  • Numerical Methods for Differential Equations MATH 5810
  • Analysis I MATH 4140
  • Introduction to Mathematical Logic MATH 4430X
  • Partial Differential Equations MATH 3850
  • Theory of Linear Algebra, honors MATH 3170
Computing and machine learning
  • Concepts and Applications of Machine Learning, honors DS 3030
  • Design and Analysis of Algorithms COM S 3110
  • Software Development Practices COM S 3090
Research

Research experience

Langevin Sampling and Diffusion Estimation for Molecular Systems

  • Built an end-to-end Python simulation pipeline implementing MALA and projected Euler–Maruyama samplers for overdamped Langevin dynamics, staged from a one-dimensional periodic potential to a solvated ion in a three-dimensional Lennard-Jones bath; estimated self-diffusion and mobility coefficients from long-run trajectory data.
  • Derived the closed-form Lifson–Jackson diffusivity D = 1/I₀(1)² ≈ 0.62386 as an exact benchmark and validated the estimators against it before scaling to three dimensions, where the pipeline reproduces published solvated-ion results.
  • Author weekly technical reports, including a companion note supporting the group's CLT-based error analysis.

Large Deviation Theory for Rare-Event Analysis in Queueing Systems

  • Developed the first importance-sampling methodology for rare-event estimation in the mean-field join-the-shortest-queue model, deriving state-dependent exponential tilting from the Budhiraja–Friedlander–Wu (2021) large-deviation rate function; stated a logarithmic-efficiency conjecture with a proof sketch via the Dupuis–Wang subsolution framework.
  • Built a reproducible, performance-tuned Python simulation framework (Xoshiro256** with deterministic seeding) and validated it across 27 configurations (n ≤ 100, ρ ∈ {0.7, 0.9, 0.95}): estimates agree with crude Monte Carlo within 95% confidence intervals wherever crude Monte Carlo succeeds, and extend to probabilities as small as 10⁻³¹, where it fails entirely.
  • Designed and benchmarked two heuristic tilting families (rate-reversal, level-weighted) with pilot-run selection; diagnosed over-tilting at extreme rarity through likelihood-ratio calibration statistics, motivating the ongoing computation of the optimal state-dependent tilt.

Stability Analysis of Cell-Average Neural Networks for PDEs

  • Adapting von Neumann stability analysis to neural-network PDE solvers: deriving the linearized amplification factor of the learned time-stepping scheme and characterizing how perturbations propagate through it, toward stability conditions on the trained weights analogous to those of classical finite-difference methods.
  • Reimplementing the group's solver in Python to enable systematic stability experiments across schemes and step sizes.
Manuscripts

Manuscripts in preparation

  1. Large-Deviation-Guided Importance Sampling for Rare Events in the Mean-Field Join-the-Shortest-Queue Model
  2. Langevin Dynamics and Transport-Coefficient Estimation for Molecular Systems
  3. Stability Certificates for Cell-Average-Based Neural Network Solvers: von Neumann Analysis of Trained Flux Networks
Awards

Awards and honors

  • Dean's High Impact Undergraduate Research Award, Iowa State University, 2026.
  • Schillmoeller Family Scholarship in Statistics.
  • Mead Fund Award in Statistics.
  • James L. Cornette Mathematics and Interdisciplinary Studies Award.
  • Dean's List, every semester.
Service

Leadership and service

  • Treasurer, Iowa State Data Science Club, 2025 – present. Secured $60,000+ in funding supporting 130+ members.
  • Research Ambassador, Iowa State University Honors Program.
  • FAA Private Pilot Certificate, high-performance endorsement, 2025. Volunteer pilot and recruitment advisor, Pilots N Paws animal-rescue flights.
Skills

Technical skills

Languages and tools
PyTorch and CUDA for neural-network and GPU-accelerated simulation; Python (NumPy, SciPy, pandas, scikit-learn, matplotlib); R; LaTeX; git.
Methods
Measure-theoretic probability and stochastic differential equations; MCMC (MALA, Metropolis–Hastings) and importance sampling; large-deviation asymptotics; numerical analysis of time-stepping schemes and von Neumann stability; reproducible computation with recorded seeds and parameter sidecars, so every figure regenerates from saved results.