Projects
Research Topics:
Frameworks
- Multiscale Evolution (muse): code for our papers on (i) Evolutionary Implications of Multiscale Intelligence, and (ii) Aging as a Loss of Goal-Directedness, i.e., (i) an implementation for evolutionary training of (recurrent gene regulatory) developmental Neural Cellular Automata (NCAs), and (ii) for investigating aging and rejuvenation strategies in evolved NCAs that are deployed far beyond their target life-spans.
- Conditional Diffusion Evolution (condevo): a deep-learning based approach to use iteratively refined diffusion models to generate high-quality, well adapted offspring in an ongoing evolutionary search. Classifier-free guidance techniques allow guiding the generative denoising towards solutions with target traits via conditional sampling in parameter space, for phenotypical traits, and population-wide features.
- Diffusion Evolution (diffevo): a model-free approach of using diffusion models as evolutionary algorithm, allowing for leveraging advancements in diffusion models for evolutionary algorithm tasks, including accelerated sampling and latent space diffusion.
- mindcraft: a framework for cognitive neuroevolution
- bio-feedback: a framework to synchronously gather, process, and control bio-sensory data
- data-monitor: an interactive (non-blocking)
matplotlib-based time-series visualization tool
- gempy:
Python implementations of different generative models (maximum entropy method -
NumPy & Numba, Variational-Autoencoder - PyTorch)
- cthru: a software to minimize reflection in video data of an optical multi-polarization filter system
- atuin: a Python-based framework for evolutionary optimization of nested data structures - developed during my Ph.D. (available on demand)
Coding Challenges
Applications and Hobby
- Inkstrument: an AI-based music controller (WIP)
- Raspberry Pi brewery (putting CraftBeerPi by Manuel Fritsch to use)