https://www2.econ.iastate.edu/tesfatsi/DeepLearningInNeuralNetworksOverview.JSchmidhuber2015.pdf

Deep learning in neural networks

I. the origin of species by means of natural selection

Policy search methods can allow robots to learn control policies for a wide range of tasks, but practical applications of policy search often require hand-engineered components for perception, state estimation, and low-level control. In this paper, we aim to answer the following question: does training the perception and control systems jointly end-to-end provide better performance than training each component separately? To this end, we develop a method that can be used to learn policies that map raw image observations directly to torques at the robot's motors. The policies are represented by deep convolutional neural networks (CNNs) with 92,000 parameters, and are trained using a guided policy search method, which transforms policy search into supervised learning, with supervision provided by a simple trajectory-centric reinforcement learning method. We evaluate our method on a range of real-world manipulation tasks that require close coordination between vision and control, such as screwing a cap onto a bottle, and present simulated comparisons to a range of prior policy search methods.

/pdf/end-to-end-training-of-deep-visuomotor-policies-1vqd5uo50d.pdf

End-to-end training of deep visuomotor policies

Deep reinforcement learning (DRL) is poised to revolutionize the field of artificial intelligence (AI) and represents a step toward building autonomous systems with a higherlevel understanding of the visual world. Currently, deep learning is enabling reinforcement learning (RL) to scale to problems that were previously intractable, such as learning to play video games directly from pixels. DRL algorithms are also applied to robotics, allowing control policies for robots to be learned directly from camera inputs in the real world. In this survey, we begin with an introduction to the general field of RL, then progress to the main streams of value-based and policy-based methods. Our survey will cover central algorithms in deep RL, including the deep Q-network (DQN), trust region policy optimization (TRPO), and asynchronous advantage actor critic. In parallel, we highlight the unique advantages of deep neural networks, focusing on visual understanding via RL. To conclude, we describe several current areas of research within the field.

Deep Reinforcement Learning: A Brief Survey

Deep reinforcement learning is poised to revolutionise the field of AI and represents a step towards building autonomous systems with a higher level understanding of the visual world. Currently, deep learning is enabling reinforcement learning to scale to problems that were previously intractable, such as learning to play video games directly from pixels. Deep reinforcement learning algorithms are also applied to robotics, allowing control policies for robots to be learned directly from camera inputs in the real world. In this survey, we begin with an introduction to the general field of reinforcement learning, then progress to the main streams of value-based and policy-based methods. Our survey will cover central algorithms in deep reinforcement learning, including the deep $Q$-network, trust region policy optimisation, and asynchronous advantage actor-critic. In parallel, we highlight the unique advantages of deep neural networks, focusing on visual understanding via reinforcement learning. To conclude, we describe several current areas of research within the field.

A brief survey of deep reinforcement learning

The idea of using evolutionary computation to train artificial neural networks, or neuroevolution (NE), for reinforcement learning (RL) tasks has now been around for over 20 years. However, as RL tasks become more challenging, the networks required become larger, as do their genomes. But, scaling NE to large nets (i.e. tens of thousands of weights) is infeasible using direct encodings that map genes one-to-one to network components. In this paper, we scale-up our compressed network encoding where network weight matrices are represented indirectly as a set of Fourier-type coefficients, to tasks that require very-large networks due to the high-dimensionality of their input space. The approach is demonstrated successfully on two reinforcement learning tasks in which the control networks receive visual input: (1) a vision-based version of the octopus control task requiring networks with over 3 thousand weights, and (2) a version of the TORCS driving game where networks with over 1 million weights are evolved to drive a car around a track using video images from the driver's perspective.

/pdf/evolving-large-scale-neural-networks-for-vision-based-4y4a4v9vqt.pdf

Evolving large-scale neural networks for vision-based reinforcement learning

The idea of evolving novel rather than fit solutions has recently been offered as a way to automatically discover the kind of complex solutions that exhibit truly intelligent behavior. So far, novelty search has only been studied in the context of problems where the number of possible "different" solutions has been limited. In this paper, we show, using a task with a much larger solution space, that selecting for novelty alone does not offer an advantage over fitness-based selection. In addition, we examine how the idea of novelty search can be used to sustain diversity and improve the performance of standard, fitness-based search.

When novelty is not enough

Deep reinforcement learning, applied to vision-based problems like Atari games, maps pixels directly to actions; internally, the deep neural network bears the responsibility of both extracting useful information and making decisions based on it. By separating the image processing from decision-making, one could better understand the complexity of each task, as well as potentially find smaller policy representations that are easier for humans to understand and may generalize better. To this end, we propose a new method for learning policies and compact state representations separately but simultaneously for policy approximation in reinforcement learning. State representations are generated by an encoder based on two novel algorithms: Increasing Dictionary Vector Quantization makes the encoder capable of growing its dictionary size over time, to address new observations as they appear in an open-ended online-learning context; Direct Residuals Sparse Coding encodes observations by disregarding reconstruction error minimization, and aiming instead for highest information inclusion. The encoder autonomously selects observations online to train on, in order to maximize code sparsity. As the dictionary size increases, the encoder produces increasingly larger inputs for the neural network: this is addressed by a variation of the Exponential Natural Evolution Strategies algorithm which adapts its probability distribution dimensionality along the run. We test our system on a selection of Atari games using tiny neural networks of only 6 to 18 neurons (depending on the game's controls). These are still capable of achieving results comparable---and occasionally superior---to state-of-the-art techniques which use two orders of magnitude more neurons.

/pdf/playing-atari-with-six-neurons-1gciswrr48.pdf

Playing Atari with Six Neurons

The TORCS racing simulator has become a standard testbed used in many recent reinforcement learning competitions, where an agent must learn to drive a car around a track using a small set of task-specific features. In this paper, large, recurrent neural networks (with over 1 million weights) are evolved to solve a much more challenging version of the task that instead uses only a stream of images from the driver’s perspective as input. Evolving such large nets is made possible by representing them in the frequency domain as a set of coefficients that are transformed into weight matrices via an inverse Fourier-type transform. To our knowledge this is the first attempt to tackle TORCS using vision, and successfully evolve a neural network controllers of this size.

http://www.fdg2013.org/program/papers/paper27_koutnik_etal.pdf

Evolving Large-Scale Neural Networks for Vision-Based TORCS

Neuroevolution, the artificial evolution of neural networks, has shown great promise on continuous reinforcement learning tasks that require memory. However, it is not yet directly applicable to realistic embedded agents using high-dimensional (e.g. raw video images) inputs, requiring very large networks. In this paper, neuroevolution is combined with an unsupervised sensory pre-processor or compressor that is trained on images generated from the environment by the population of evolving recurrent neural network controllers. The compressor not only reduces the input cardinality of the controllers, but also biases the search toward novel controllers by rewarding those controllers that discover images that it reconstructs poorly. The method is successfully demonstrated on a vision-based version of the well-known mountain car benchmark, where controllers receive only single high-dimensional visual images of the environment, from a third-person perspective, instead of the standard two-dimensional state vector which includes information about velocity.

/pdf/intrinsically-motivated-neuroevolution-for-vision-based-3gkvdaz1ie.pdf

Giuseppe Cuccu

Papers

Evolving large-scale neural networks for vision-based reinforcement learning

When novelty is not enough

Playing Atari with Six Neurons

Evolving Large-Scale Neural Networks for Vision-Based TORCS

Intrinsically motivated neuroevolution for vision-based reinforcement learning