Skip to content
View RadoslavJochman's full-sized avatar

Block or report RadoslavJochman

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
RadoslavJochman/README.md

Radoslav Jochman

MSc student in Computational Biology and Bioinformatics at ETH Zürich (D-BSSE), Basel. I build Bayesian statistical models for biological data, across cancer genomics and neuroscience, and I care about the engineering as much as the statistics: a model that samples cleanly and a codebase fast enough to actually test it in.

My research has covered hierarchical Bayesian inference on tumour phylogenies, in the Computational Biology Group (Beerenwinkel Lab) at D-BSSE, and decoding functional structure from visual cortex population activity, as my bachelor thesis at Charles University in Prague.

Most of my day to day work is in Python. I'm drawn to neural networks and deep learning, both as models to build with and as things to understand, and my projects span probabilistic modelling, reinforcement learning, and neuroevolution.

Outside that I write C++ and C#, usually to make something that learns or evolves and that you can watch doing it.


Projects

Tree-HDP: mutational signatures on tumour phylogenies A hierarchical Dirichlet process that infers mutational signature activity across tumour subclones using the phylogeny relating them, structure that standard NMF discards. Fixed signature and de novo variants in PyMC/NumPyro, benchmarked against NMF. Bayesian inference MCMC PyMC cancer genomics

Spontaneous map inference Decoding orientation preference maps from spontaneous activity in human V1, and testing whether sleep gives more stable maps than wakefulness. PCA based decoding, stability analysis across bin size and time. computational neuroscience signal processing scikit-learn

AlphaZero from scratch Self play reinforcement learning for 8×8 five in a row with the pie rule, in C#/TorchSharp. Three trained checkpoints included, so you can play it. Moving the hot path off tensors made self play about 35× faster. reinforcement learning MCTS C# TorchSharp

Evolution simulator Creatures with genome encoded neural networks evolving under selection pressure, rendered live in Qt. Modern C++20 and CMake. neuroevolution C++ Qt6


Tools

Python C++ C# SQL Bash · PyMC NumPyro PyTorch NumPy pandas scikit-learn · Git Docker LaTeX Linux CMake

Slovak (native) · Czech (fluent) · English (C1) · German (B1)


📫 rdjchmn@gmail.com

Popular repositories Loading

  1. RadoslavJochman RadoslavJochman Public

    Config files for my GitHub profile.

  2. Bachelor-Thesis-Code-Spontaneous-Map-Inference Bachelor-Thesis-Code-Spontaneous-Map-Inference Public

    Code for my BSc thesis: decoding orientation preference maps from spontaneous activity in human V1. Compares sleep against wakefulness and finds sleep-derived maps markedly more stable — a stimulus…

    Python

  3. Mutational_signatures_HDP Mutational_signatures_HDP Public

    Bayesian hierarchical Dirichlet process inferring mutational-signature activity across tumour subclones using the phylogeny relating them. PyMC/NumPyro.

    Python

  4. evolution-simulator evolution-simulator Public

    Neuro-evolution simulator in modern C++/Qt: creatures with genome-encoded neural networks evolve under selection.

    C++

  5. alphazero-tictactoe alphazero-tictactoe Public

    Description: "AlphaZero from scratch in C#: self-play RL for 8×8 five-in-a-row with the pie rule. Play against three trained models.

    C#