Reinforcement learning (RL) has been very successful in recent years but, limited by its sample inefficiency, often requires large computational resources. While new methods are being investigated to increase the efficiency of RL algorithms it is critical to enable training at scale, yet using a code-base flexible enough to allow for method experimentation. Here, we present NAPPO, a pytorch-based library for RL which provides scalable proximal policy optimization (PPO) implementations in a simple, modular package. We validate it by replicating previous results on Mujoco and Atari environments. Furthermore, we provide insights on how a variety of distributed training schemes with synchronous and asynchronous communication patterns perform. Finally we showcase NAPPO by obtaining the highest to-date test performance on the Obstacle Tower Unity3D challenge environment. The full source code is available.

What goals should a multi-goal reinforcement learning agent pursue during training in long-horizon tasks? When the desired (test time) goal distribution is too distant to offer a useful learning signal, we argue that the agent should not pursue unobtainable goals. Instead, it should set its own intrinsic goals that maximize the entropy of the historical achieved goal distribution. We propose to optimize this objective by having the agent pursue past achieved goals in sparsely explored areas of the goal space, which focuses exploration on the frontier of the achievable goal set. We show that our strategy achieves an order of magnitude better sample efficiency than the prior state of the art on long-horizon multi-goal tasks including maze navigation and block stacking.

Deep model-based Reinforcement Learning (RL) has the potential to substantially improve the sample-efficiency of deep RL. While various challenges have long held it back, a number of papers have recently come out reporting success with deep model-based methods. This is a great development, but the lack of a consistent metric to evaluate such methods makes it difficult to compare various approaches. For example, the common single-task sample-efficiency metric conflates improvements due to model-based learning with various other aspects, such as representation learning, making it difficult to assess true progress on model-based RL. To address this, we introduce an experimental setup to evaluate model-based behavior of RL methods, inspired by work from neuroscience on detecting model-based behavior in humans and animals. Our metric based on this setup, the Local Change Adaptation (LoCA) regret, measures how quickly an RL method adapts to a local change in the environment. Our metric can identify model-based behavior, even if the method uses a poor representation and provides insight in how close a method's behavior is from optimal model-based behavior. We use our setup to evaluate the model-based behavior of MuZero on a variation of the classic Mountain Car task.

Imitation learning is a powerful family of techniques for learning sensorimotor coordination in immersive environments. We apply imitation learning to attain state-of-the-art performance on hard exploration problems in the Minecraft environment. We report experiments that highlight the influence of network architecture, loss function, and data augmentation. An early version of our approach reached second place in the MineRL competition at NeurIPS 2019. Here we report stronger results that can be used as a starting point for future competition entries and related research. Our code is available at https://github.com/amiranas/minerl_imitation_learning.

Standard RL algorithms assume fixed environment dynamics and require a significant amount of interaction to adapt to new environments. We introduce Policy-Dynamics Value Functions (PD-VF), a novel approach for rapidly adapting to dynamics different from those previously seen in training. PD-VF explicitly estimates the cumulative reward in a space of policies and environments. An ensemble of conventional RL policies is used to gather experience on training environments, from which embeddings of both policies and environments can be learned. Then, a value function conditioned on both embeddings is trained. At test time, a few actions are sufficient to infer the environment embedding, enabling a policy to be selected by maximizing the learned value function (which requires no additional environment interaction). We show that our method can rapidly adapt to new dynamics on a set of MuJoCo domains. Code available at https://github.com/rraileanu/policy-dynamics-value-functions.

Radio Frequency powered Cognitive Radio Networks (RF-CRN) are likely to be the eyes and ears of upcoming modern networks such as Internet of Things (IoT), requiring increased decentralization and autonomous operation. To be considered autonomous, the RF-powered network entities need to make decisions locally to maximize the network throughput under the uncertainty of any network environment. However, in complex and large-scale networks, the state and action spaces are usually large, and existing Tabular Reinforcement Learning technique is unable to find the optimal state-action policy quickly. In this paper, deep reinforcement learning is proposed to overcome the mentioned shortcomings and allow a wireless gateway to derive an optimal policy to maximize network throughput. When benchmarked against advanced DQN techniques, our proposed DQN configuration offers performance speedup of up to 1.8x with good overall performance.

We investigate whether Jacobi preconditioning, accounting for the bootstrap term in temporal difference (TD) learning, can help boost performance of adaptive optimizers. Our method, TDprop, computes a per parameter learning rate based on the diagonal preconditioning of the TD update rule. We show how this can be used in both $n$-step returns and TD($\lambda$). Our theoretical findings demonstrate that including this additional preconditioning information is, surprisingly, comparable to normal semi-gradient TD if the optimal learning rate is found for both via a hyperparameter search. In Deep RL experiments using Expected SARSA, TDprop meets or exceeds the performance of Adam in all tested games under near-optimal learning rates, but a well-tuned SGD can yield similar improvements -- matching our theory. Our findings suggest that Jacobi preconditioning may improve upon typical adaptive optimization methods in Deep RL, but despite incorporating additional information from the TD bootstrap term, may not always be better than SGD.

Applying Deep Reinforcement Learning (DRL) to complex tasks in the field of robotics has proven to be very successful in the recent years. However, most of the publications focus either on applying it to a task in simulation or to a task in a real world setup. Although there are great examples of combining the two worlds with the help of transfer learning, it often requires a lot of additional work and fine-tuning to make the setup work effectively. In order to increase the use of DRL with real robots and reduce the gap between simulation and real world robotics, we propose an open source toolkit: robo-gym. We demonstrate a unified setup for simulation and real environments which enables a seamless transfer from training in simulation to application on the robot. We showcase the capabilities and the effectiveness of the framework with two real world applications featuring industrial robots: a mobile robot and a robot arm. The distributed capabilities of the framework enable several advantages like using distributed algorithms, separating the workload of simulation and training on different physical machines as well as enabling the future opportunity to train in simulation and real world at the same time. Finally we offer an overview and comparison of robo-gym with other frequently used state-of-the-art DRL frameworks.

Machine learning is an application of artificial intelligence which refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. The term may also be applied to any machine that exhibits traits associated with a human mind such as learning and problem-solving. SIMPLY put Machine Learning is the science of getting computers to learn and act like humans do, and improve their learning over time in autonomous fashion, by feeding them data and information in the form of observations and real-world interactions. This shows the ideal objective or ultimate aim of machine learning, which may make a lot of people confused between Artificial intelligence and Machine Learning however, there are some distinct differences readers should recognize as well. Machine learning is just one method for achieving artificial intelligence.

Paraphrasing is expressing the meaning of an input sentence in different wording while maintaining fluency (i.e., grammatical and syntactical correctness). Most existing work on paraphrasing use supervised models that are limited to specific domains (e.g., image captions). Such models can neither be straightforwardly transferred to other domains nor generalize well, and creating labeled training data for new domains is expensive and laborious. The need for paraphrasing across different domains and the scarcity of labeled training data in many such domains call for exploring unsupervised paraphrase generation methods. We propose Progressive Unsupervised Paraphrasing (PUP): a novel unsupervised paraphrase generation method based on deep reinforcement learning (DRL). PUP uses a variational autoencoder (trained using a non-parallel corpus) to generate a seed paraphrase that warm-starts the DRL model. Then, PUP progressively tunes the seed paraphrase guided by our novel reward function which combines semantic adequacy, language fluency, and expression diversity measures to quantify the quality of the generated paraphrases in each iteration without needing parallel sentences. Our extensive experimental evaluation shows that PUP outperforms unsupervised state-of-the-art paraphrasing techniques in terms of both automatic metrics and user studies on four real datasets. We also show that PUP outperforms domain-adapted supervised algorithms on several datasets. Our evaluation also shows that PUP achieves a great trade-off between semantic similarity and diversity of expression.