Severi Rissanen

Severi Rissanen

PhD student in Machine Learning

Aalto University

Biography

I am a PhD student in the Department of Computer Science at Aalto University, advised by Arno Solin and Markus Heinonen. My research interests have included diffusion generative models, (physics-inspired) inductive biases in diffusion models, adding inference-time controls to diffusion models and diffusion models applications to scientific problems. Previously, I also worked in the intersection of deep latent variable models and causal inference. During my PhD, I spent time as a visiting student in University of Cambridge in José Miguel Hernández-Lobato’s group and did an internship at Microsoft Research AI4Science in Cambridge. I published 7 papers in NeurIPS/ICML/ICLR in my Master’s and PhD so far, with 5 of them first-author/shared-first-author. I was also one of the main organisers of ELLIS Doctoral Symposium 2023!

Education
  • PhD in Machine Learning, 2021-

    Aalto University

  • MSc in Machine Learning, 2019-2021

    Aalto University

  • BSc in Engineering Physics and Mathematics, 2016-2019

    Aalto University

Publications

(2026). Pareto-Conditioned Diffusion Models for Offline Multi-Objective Optimization. In ICLR 2026.

PDF

(2025). PriorGuide: Test-Time Prior Adaptation for Simulation-Based Inference. In ICLR 2026.

PDF

(2025). Progressive Tempering Sampler with Diffusion. In ICML 2025.

PDF

(2024). Free Hunch: Denoiser Covariance Estimation for Diffusion Models Without Extra Costs. In ICLR 2025.

PDF

(2024). Alignment is Key for Applying Diffusion Models to Retrosynthesis. In ICLR 2025.

PDF

Recent Talks

Other / blog posts

*
Transcribing my grandma’s memoirs with LLMs
This post details my attempts at automatically transcribing my grandma’s handwritten memoirs into somewhat modern, easily readable Finnish.
Transcribing my grandma's memoirs with LLMs
Old School Projects
A couple of selected projects I did before starting my PhD, based on some measure of interestingness.
Old School Projects