A universal supervisor controller and ER suite for Webots that can be adapted to any wheeled robot morphology with ease. The project is also setup to allow for easy Reinforcement Learning experimentation with some select algorithms (CMA-ES, Novlty Search, MAP-Elites) and neural networks (fixed and recurrent).

Python implementation of Regulated Evolution Strategies with Covariance Matrix Adaption for continuous "Black-Box" optimization problems.

Master project

(EvoApps2022) "Towards a Principled Learning Rate Adaptation for Natural Evolution Strategies"

Official implementation of "Approximating Gradients for Differentiable Quality Diversity in Reinforcement Learning"

LLaMA-ES: Tuning CMA-ES hyperparameters using LLaMA3 for optimized numerical function performance.

(GECCO2023 Best Paper Nomination & ACM TELO) CMA-ES with Learning Rate Adaptation

Reproduce the results of "Neuroevolution of Self-Interpretable Agents" paper

(CEC2022) Fast Moving Natural Evolution Strategy for High-Dimensional Problems

Convert images into low poly, using an optimizer

(GECCO 2022) CMA-ES with Margin: Lower-Bounding Marginal Probability for Mixed-Integer Black-Box Optimization

A new version of world models using Echo-state networks and random weight-fixed CNNs

Deep learning and evolutionary algorithms for identification of aerodynamic parameters

Code for: Transforming and Projecting Images into Class-conditional Generative Networks

Distributed implementation of popular evolutionary methods

Official implementation of the MM'21 paper "Constrained Graphic Layout Generation via Latent Optimization" (LayoutGAN++, CLG-LO, and Layout evaluation)

A bare-bones Python library for quality diversity optimization.

Python library for CMA Evolution Strategy.