ml ・ design
Permutation-Invariant Neural Networks for Reinforcement Learning
November 18, 2021
Reinforcement learning agents typically perform poorly if provided with inputs that were not clearly defined in training. A new approach enables RL agents to perform well, even when subject to corrupt, incomplete, or shuffled inputs.
Modern Evolution Strategies for Creativity:Fitting Concrete Images and Abstract Concepts
September 21, 2021
“A drawing of a cat”
CLIP + ES + Triangles
Neuroevolution of Self-Interpretable Agents
March 18, 2020
Agents with a self-attention “bottleneck” not only can solve these tasks from pixel inputs with only 4000 parameters, but they are also better at generalization.
Learning to Predict Without Looking Ahead
October 29, 2019
Rather than hardcoding forward prediction, we try to get agents to learn that they need to predict the future.
Weight Agnostic Neural Networks
June 12, 2019
We search for neural network architectures that can already perform various tasks even when they use random weight values.