Hirofumi Shiba

司馬博文
1st year Ph.D. student in Statistics
ORCID Github Twitter Email

Institute of Statistical Mathematics (ISM), Tokyo, Japan
Graduate University for Advanced Studies (SOKENDAI)

Feel free to call me Hiro.
I work chiefly on Bayesian computation, supervised by Kengo Kamatani.

Interests

  1. Sampling methods, e.g., MCMC, SMC, Diffusion models, etc.
  2. Optimization methods, e.g. SGD, Evolution strategies, etc.
  3. Bayesian modelling, especially in political science & biostatistics

My ultimate interest lies in the dynamics of computational algorithms, such as Monte Carlo and optimization algorithms, and their resemblance to natural phenomena.

Specifically, I recently work on emerging sampling algorithms, such as Piecewise Deterministic Monte Carlo, Hamiltonian Monte Carlo, and Denoising Diffusion Models.

Before pursuing my Ph.D., I was trained in mathematics at the University of Tokyo, with an emphasis on Stochastic Processes and Functional Analysis.

My surname Shiba (司馬) has different kanji from the Shiba (柴) dog, although they are equally charming.

Software

I maintain the Julia package PDMPFlux.jl, and also contribute to the R package YUIMA.