Nikita Morozov
website picture

Hello, I am Nikita! I am a first year PhD student at HSE University supervised by Prof. Dmitry Vetrov and a part of Centre of Deep Learning and Bayesian Methods. My research is focused on Generative Flow Networks (GFlowNets) and their connections to Reinforcement Learning. I am working towards creating new efficient GFlowNet training algorithms, as well as expanding the theoretical understanding of these models. Generally speaking, my research interests include generative modeling and probabilistic methods in deep learning, as well as their applications in the sciences.

I did a research internship at EPFL MLBIO under the supervision of Prof. Maria Brbić, where I worked with Yist Yu on geometric diffusion models for single-cell data. I have also previously worked on novel view synthesis and text-to-3D generative models with Kirill Struminsky. Prior to joining Bayesian Methods Research Group, I was a software engineer intern at Yandex, where I worked on refining YTsaurus platform.

I have a background in competitive programming, I participated in 2024 ICPC World Finals and you can find me on codeforces.

Email / CV (Last update: July 2024) / GitHub / Google Scholar / HSE webpage

 

Publications

  • Generating complex tissue structures from gene expressions with LUNA
    Tingyang Yu, Chanakya Ekbote, Nikita Morozov, Stéphane d’Ascoli, Jiashuo Fan, Pascal Frossard, Maria Brbić
    bioRxiv TBA / code TBA
    Preprint 2024

  • Optimizing Backward Policies in GFlowNets via Trajectory Likelihood Maximization
    Timofei Gritsaev, Nikita Morozov, Sergey Samsonov, Daniil Tiapkin
    arXiv / code
    Preprint 2024

  • Improving GFlowNets with Monte Carlo Tree Search
    Nikita Morozov, Daniil Tiapkin, Sergey Samsonov, Alexey Naumov, Dmitry Vetrov
    arXiv / code TBA
    ICML 2024 Workshop on Structured Probabilistic Inference & Generative Modeling

  • Generative Flow Networks as Entropy-Regularized RL
    Daniil Tiapkin*, Nikita Morozov*, Alexey Naumov, Dmitry Vetrov
    arXiv / code
    AISTATS 2024 (Oral)

  • Differentiable Rendering with Reparameterized Volume Sampling
    Nikita Morozov, Denis Rakitin, Oleg Desheulin, Dmitry Vetrov, Kirill Struminsky
    arXiv / code
    AISTATS 2024
    Short version appeared in ICLR 2023 Workshop on Neural Fields

  • Weight Averaging Improves Knowledge Distillation under Domain Shift
    Valeriy Berezovskiy, Nikita Morozov
    arXiv / code
    ICCV 2023 Workshop on Out-of-Distribution Generalization in Computer Vision