Gradient descent or its variants are popular in training neural networks. However, in deep Q-learning with neural network approximation, a type of reinforcement learning, gradient descent (also known as Residual Gradient (RG)) is barely used to solve Bellman residual minimization problem. On the contrary, Temporal Difference (TD), an incomplete gradient descent method prevails. In this work, we perform extensive experiments to show that TD outperforms RG, that is, when the training leads to a small Bellman residual error, the solution found by TD has a better policy and is more robust against the perturbation of neural network parameters. We further use experiments to reveal a key difference between reinforcement learning and supervised learning, that is, a small Bellman residual error can correspond to a bad policy in reinforcement learning while the test loss function in supervised learning is a standard index to indicate the performance. We also empirically examine that the missing term in TD is a key reason why RG performs badly. Our work shows that the performance of a deep Q-learning solution is closely related to the training dynamics and how an incomplete gradient descent method can find a good policy is interesting for future study.

In the previous article, I shared the paper(you can follow the link below to recap!!!): Weighted QMIX: Expanding Monotonic Value Function Factorization for Deep Multi-Agent Reinforcement Learning, which argues that the previous approaches, such as VDN and QMIX, can only factorize a little group of tasks, and proposed a new framework to overcome the issue. In this article, I gonna share another way to factorize any factorizable task, which is called QPLEX!!! In most of the multi-agent approaches, we tend to explore the popular paradigm of centralized training with decentralized execution(CTDE). In this paradigm, individual-Global-Max(IGM) principle plays an important role. However, lots of the methods tend to relax the IGM consistency so that they can achieve scalability.

Consider a target being tracked by a cognitive radar network. If the target can intercept some radar network emissions, how can it detect coordination among the radars? By 'coordination' we mean that the radar emissions satisfy Pareto optimality with respect to multi-objective optimization over each radar's utility. This paper provides a novel multi-objective inverse reinforcement learning approach which allows for both detection of such Pareto optimal ('coordinating') behavior and subsequent reconstruction of each radar's utility function, given a finite dataset of radar network emissions. The method for accomplishing this is derived from the micro-economic setting of Revealed Preferences, and also applies to more general problems of inverse detection and learning of multi-objective optimizing systems.

Reinforcement learning has seen increasing applications in real-world contexts over the past few years. However, physical environments are often imperfect and policies that perform well in simulation might not achieve the same performance when applied elsewhere. A common approach to combat this is to train agents in the presence of an adversary. An adversary acts to destabilise the agent, which learns a more robust policy and can better handle realistic conditions. Many real-world applications of reinforcement learning also make use of goal-conditioning: this is particularly useful in the context of robotics, as it allows the agent to act differently, depending on which goal is selected. Here, we focus on the problem of goal-conditioned learning in the presence of an adversary. We first present DigitFlip and CLEVR-Play, two novel goal-conditioned environments that support acting against an adversary. Next, we propose EHER and CHER -- two HER-based algorithms for goal-conditioned learning -- and evaluate their performance. Finally, we unify the two threads and introduce IGOAL: a novel framework for goal-conditioned learning in the presence of an adversary. Experimental results show that combining IGOAL with EHER allows agents to significantly outperform existing approaches, when acting against both random and competent adversaries.

Agents can base decisions made using reinforcement learning (RL) on a reward function. The selection of values for the learning algorithm parameters can, nevertheless, have a substantial impact on the overall learning process. In order to discover values for the learning parameters that are close to optimal, we extended our previously proposed genetic algorithm-based Deep Deterministic Policy Gradient and Hindsight Experience Replay approach (referred to as GA+DDPG+HER) in this study. On the robotic manipulation tasks of FetchReach, FetchSlide, FetchPush, FetchPick&Place, and DoorOpening, we applied the GA+DDPG+HER methodology. Our technique GA+DDPG+HER was also used in the AuboReach environment with a few adjustments. Our experimental analysis demonstrates that our method produces performance that is noticeably better and occurs faster than the original algorithm. We also offer proof that GA+DDPG+HER beat the current approaches. The final results support our assertion and offer sufficient proof that automating the parameter tuning procedure is crucial and does cut down learning time by as much as 57%.

This paper tackles the problem of how to pre-train a model and make it generally reusable backbones for downstream task learning. In pre-training, we propose a method that builds an agent-environment interaction model by learning domain invariant successor features from the agent's vast experiences covering various tasks, then discretize them into behavior prototypes which result in an embodied set structure. To make the model generally reusable for downstream task learning, we propose (1) embodied feature projection that retains previous knowledge by projecting the new task's observation-action pair to the embodied set structure and (2) projected Bellman updates which add learning plasticity for the new task setting. We provide preliminary results that show downstream task learning based on a pre-trained embodied set structure can handle unseen changes in task objectives, environmental dynamics and sensor modalities.

In recent years, many applications have deployed incentive mechanisms to promote users' attention and engagement. Most incentive mechanisms determine specific incentive values based on users' attributes (e.g., preferences), while such information is unavailable in many real-world applications. Meanwhile, due to budget restrictions, realizing successful incentivization for all users can be challenging to complete. In this light, we consider leveraging social influence to maximize the incentivization result. We can directly incentivize influential users to affect more users, so the cost of incentivizing these users can be decreased. However, identifying influential users in a social network requires complete information about influence strength among users, which is impractical to acquire in real-world situations. In this research, we propose an end-to-end reinforcement learning-based framework, called Geometric Actor-Critic (GAC), to tackle the abovementioned problem. The proposed approach can realize effective incentive allocation without having prior knowledge about users' attributes. Three real-world social network datasets have been adopted in the experiments to evaluate the performance of GAC. The experimental results indicate that GAC can learn and apply effective incentive allocation policies in unknown social networks and outperform existing incentive allocation approaches.

Studies have shown evolution strategies (ES) to be a promising approach for reinforcement learning (RL) with deep neural networks. However, the issue of high sample complexity persists in applications of ES to deep RL over long horizons. This paper is the first to address the shortcoming of today's methods via a novel neuroevolutionary multitasking (NuEMT) algorithm, designed to transfer information from a set of auxiliary tasks (of short episode length) to the target (full length) RL task at hand. The auxiliary tasks, extracted from the target, allow an agent to update and quickly evaluate policies on shorter time horizons. The evolved skills are then transferred to guide the longer and harder task towards an optimal policy. We demonstrate that the NuEMT algorithm achieves data-efficient evolutionary RL, reducing expensive agent-environment interaction data requirements. Our key algorithmic contribution in this setting is to introduce, for the first time, a multitask skills transfer mechanism based on the statistical importance sampling technique. In addition, an adaptive resource allocation strategy is utilized to assign computational resources to auxiliary tasks based on their gleaned usefulness. Experiments on a range of continuous control tasks from the OpenAI Gym confirm that our proposed algorithm is efficient compared to recent ES baselines.

We introduce Air Learning, an open-source simulator, and a gym environment for deep reinforcement learning research on resource-constrained aerial robots. Equipped with domain randomization, Air Learning exposes a UAV agent to a diverse set of challenging scenarios. We seed the toolset with point-to-point obstacle avoidance tasks in three different environments and Deep Q Networks (DQN) and Proximal Policy Optimization (PPO) trainers. Air Learning assesses the policies' performance under various quality-of-flight (QoF) metrics, such as the energy consumed, endurance, and the average trajectory length, on resource-constrained embedded platforms like a Raspberry Pi. We find that the trajectories on an embedded Ras-Pi are vastly different from those predicted on a high-end desktop system, resulting in up to 40% longer trajectories in one of the environments. To understand the source of such discrepancies, we use Air Learning to artificially degrade high-end desktop performance to mimic what happens on a low-end embedded system. We then propose a mitigation technique that uses the hardware-in-the-loop to determine the latency distribution of running the policy on the target platform (onboard compute on the aerial robot). A randomly sampled latency from the latency distribution is then added as an artificial delay within the training loop. Training the policy with artificial delays allows us to minimize the hardware gap (discrepancy in the flight time metric reduced from 37.73% to 0.5%). Thus, Air Learning with hardware-in-the-loop characterizes those differences and exposes how the onboard compute's choice affects the aerial robot's performance. We also conduct reliability studies to assess the effect of sensor failures on the learned policies. All put together, Air Learning enables a broad class of deep RL research on UAVs. The source code is available at:http://bit.ly/2JNAVb6.

This paper explores the problem of simultaneously learning a value function and policy in deep actor-critic reinforcement learning models. We find that the common practice of learning these functions jointly is sub-optimal, due to an order-of-magnitude difference in noise levels between these two tasks. Instead, we show that learning these tasks independently, but with a constrained distillation phase, significantly improves performance. Furthermore, we find that the policy gradient noise levels can be decreased by using a lower \textit{variance} return estimate. Whereas, the value learning noise level decreases with a lower \textit{bias} estimate. Together these insights inform an extension to Proximal Policy Optimization we call \textit{Dual Network Architecture} (DNA), which significantly outperforms its predecessor. DNA also exceeds the performance of the popular Rainbow DQN algorithm on four of the five environments tested, even under more difficult stochastic control settings.