Evolving deep unsupervised convolutional networks for vision-based reinforcement learning

By Jan Koutník, J. Urgen Schmidhuber, Faustino J. Gomez. 2014

115 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Reinforcement Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Reinforcement Learning

Citations in the corpus

Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation (2016, background): “Recently, non-linear function approximators coupled with reinforcement learning [14, 16, 23] have made it possible to learn abstractions over high-dimensional state spaces, but the task of exploration with sparse feedback still remains a major challenge.”
ViZDoom: A Doom-based AI research platform for visual reinforcement learning (2016, background): “Other works in this area include teaching mobile robots with visual-based Qlearning [9], learning policies with deep auto-encoders and batch-mode algorithms [18], neuroevolution for a vision-based version of the mountain car problem [5], and compressed neuroevolution with recurrent neural networks f”
Path integral guided policy search (2016, methodology): “In [31], visual and control layers of a racing video game policy are learned separately using neuroevolution.”
Hierarchical recurrent neural network for skeleton based action recognition (2015, background): “, intelligent video surveillance, robot vision, human-computer interaction, game control, and so on [15, 36].”
Autonomous braking system via deep reinforcement learning (2017, methodology): “Recently, the DRL is applied to design control systems for autonomous driving in [11], [12].”
Evolving Deep LSTM-based Memory Networks using an Information Maximization Objective (2016, background): “Koutnik et al. [14] used a similar diversity objective (during unsupervised training) to train Convolutional Networks for simulated car racing.”
End-to-End Goal-Driven Web Navigation (2016, background): “) In much of these works, target tasks were computer games such as Atari games [8] and racing car game [6].”
WebNav: A New Large-Scale Task for Natural Language based Sequential Decision Making (2016, background): “, 2015) and racing car game (Koutnı́k et al., 2014).”
Online Evolution of Deep Convolutional Network for Vision-Based Reinforcement Learning (2014, methodology): “In this paper we extend the approach in [16].”
A Survey of End-to-End Driving: Architectures and Training Methods (2020, background): “[59] learned to drive based on image inputs in the TORCS game environment [60] via neuroevolution.”
Discovering Gated Recurrent Neural Network Architectures (2018, background): “Koutnik et al. Koutnik et al. (2014) used a similar diversity objective (during unsupervised training) to train Convolutional Networks for simulated car racing.”
Optimizing deep learning hyper-parameters through an evolutionary algorithm (2015, background): “The first focuses on replacing stochastic gradient descent and backpropagation with a GA [12, 13, 15, 16, 20, 21].”
Large scale evolution of convolutional neural networks using volunteer computing (2017, unknown): “Koutn´ık et al. have published a work titled Evolving Deep Unsupervised Convolutional Networks for Vision-Based Reinforcement Learning [19], however in this work the structure of the CNN used was held ﬁxed while only a small recurrent neural network controller (which takes output from the CNN) was…”