Countering Poisonous Inputs with Memetic Neuroevolution

Authors:
Julian Togelius;Tom Schaul;Jürgen Schmidhuber;Faustino Gomez
Affiliations:
IDSIA, Manno-Lugano, Switzerland 6298;IDSIA, Manno-Lugano, Switzerland 6298;IDSIA, Manno-Lugano, Switzerland 6298 and TU Munich, Garching, München, Germany 85748;IDSIA, Manno-Lugano, Switzerland 6298
Venue:
Proceedings of the 10th international conference on Parallel Problem Solving from Nature: PPSN X
Year:
2008

Citing 4
Cited 1

Evolving neural networks through augmenting topologies

Evolutionary Computation
Completely Derandomized Self-Adaptation in Evolution Strategies

Evolutionary Computation
Evolutionary reinforcement learning of artificial neural networks

International Journal of Hybrid Intelligent Systems - Hybridization of Intelligent Systems
Accelerated Neural Evolution through Cooperatively Coevolved Synapses

The Journal of Machine Learning Research

Genetic search feature selection for affective modeling: a case study on reported preferences

Proceedings of the 3rd international workshop on Affective interaction in natural environments

Quantified Score

Hi-index	0.00

Visualization

Abstract

Applied to certain problems, neuroevolution frequently gets stuck in local optima with very low fitness; in particular, this is true for some reinforcement learning problems where the input to the controller is a high-dimensional and/or ill-chosen state description. Evidently, some controller inputs are "poisonous", and their inclusion induce such local optima. Previously, we proposed the memetic climber, which evolves neural network topology and weights at different timescales, as a solution to this problem. In this paper, we further explore the memetic climber, and introduce its population-based counterpart: the memetic ES. We also explore which types of inputs are poisonous for two different reinforcement learning problems.