# Autoencoder-augmented Neuroevolution for Visual Doom Playing

arXiv: https://arxiv.org/abs/1707.03902 · HTML: https://mlsanity.com/paper/1707.03902/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Jul 2017.

## Built on by (most cited first, 8 of 8)

- [Counterfactual State Explanations for Reinforcement Learning Agents via Generative Deep Learning](https://mlsanity.com/paper/2101.12446/) (2021)
- [Evolutionary Reinforcement Learning: A Survey](https://mlsanity.com/paper/2303.04150/) (2023)
- [Unsupervised Representation Learning in Deep Reinforcement Learning: A Review](https://mlsanity.com/paper/2208.14226/) (2022)
- [Tile Embedding: A General Representation for Procedural Level Generation via Machine Learning](https://mlsanity.com/paper/2110.03181/) (2021)
- [Open-Ended Evolution for Minecraft Building Generation](https://mlsanity.com/paper/2209.03108/) (2022)
- [Will GPT-4 Run DOOM?](https://mlsanity.com/paper/2403.05468/) (2024)
- [MONCAE: Multi-Objective Neuroevolution of Convolutional Autoencoders](https://mlsanity.com/paper/2106.11914/) (2021)
- [Evolutionary Machine Learning and Games](https://mlsanity.com/paper/2311.16172/) (2023)

## Abstract (arXiv)

Neuroevolution has proven effective at many reinforcement learning tasks, but does not seem to scale well to high-dimensional controller representations, which are needed for tasks where the input is raw pixel data. We propose a novel method where we train an autoencoder to create a comparatively low-dimensional representation of the environment observation, and then use CMA-ES to train neural network controllers acting on this input data. As the behavior of the agent changes the nature of the input data, the autoencoder training progresses throughout evolution. We test this method in the VizDoom environment built on the classic FPS Doom, where it performs well on a health-pack gathering task.