Reinforcement Learning (RL) has achieved phenomenal successes in solving complex sequential decisionmaking problems (Silver et al., 2016, 2017; Levine et al., 2016; Gu et al., 2017). Most of these empirical successes are driven by policy-based (policy optimization) methods, such as policy gradient (Sutton et al., 1999), natural policy gradient (NPG) (Kakade, 2001), trust region policy optimization (TRPO) (Schulman et al., 2015), and proximal policy optimization (PPO) (Schulman et al., 2017). For example, Haarnoja et al. (2018) proposed a policy-based state-of-the-art reinforcement learning algorithm, soft actor-critic (SAC), which outperformed value-based methods in a variety of real world robotics tasks including manipulation and locomotion. In fact, Kalashnikov et al. (2018) observed that compared with value-based methods such as Q-learning, policy-based methods work better with dense reward. On the other hand, for sparse reward cases in robotics, value-based methods perform better. Motivated by this, a line of recent work (Fazel et al., 2018; Bhandari and Russo, 2019; Liu et al., 2019; Wang et al., 2019; Agarwal et al., 2021) provides global convergence guarantees for these popular policybased methods. However, to achieve this goal, they made several assumptions.

Policy gradient (PG) estimators for softmax policies are ineffective with sub-optimally saturated initialization, which happens when the density concentrates on a sub-optimal action. Sub-optimal policy saturation may arise from bad policy initialization or sudden changes in the environment that occur after the policy has already converged, and softmax PG estimators require a large number of updates to recover an effective policy. This severe issue causes high sample inefficiency and poor adaptability to new situations. To mitigate this problem, we propose a novel policy gradient estimator for softmax policies that utilizes the bias in the critic estimate and the noise present in the reward signal to escape the saturated regions of the policy parameter space. Our analysis and experiments, conducted on bandits and classical MDP benchmarking tasks, show that our estimator is more robust to policy saturation.

Renewable energy resources (RERs) have been increasingly integrated into modern power systems, especially in large-scale distribution networks (DNs). In this paper, we propose a deep reinforcement learning (DRL)-based approach to dynamically search for the optimal operation point, i.e., optimal power flow (OPF), in DNs with a high uptake of RERs. Considering uncertainties and voltage fluctuation issues caused by RERs, we formulate OPF into a multi-objective optimization (MOO) problem. To solve the MOO problem, we develop a novel DRL algorithm leveraging the graphical information of the distribution network. Specifically, we employ the state-of-the-art DRL algorithm, i.e., deep deterministic policy gradient (DDPG), to learn an optimal strategy for OPF. Since power flow reallocation in the DN is a consecutive process, where nodes are self-correlated and interrelated in temporal and spatial views, to make full use of DNs' graphical information, we develop a multi-grained attention-based spatial-temporal graph convolution network (MG-ASTGCN) for spatial-temporal graph information extraction, preparing for its sequential DDPG. We validate our proposed DRL-based approach in modified IEEE 33, 69, and 118-bus radial distribution systems (RDSs) and show that our DRL-based approach outperforms other benchmark algorithms. Our experimental results also reveal that MG-ASTGCN can significantly accelerate the DDPG training process and improve DDPG's capability in reallocating power flow for OPF. The proposed DRL-based approach also promotes DNs' stability in the presence of node faults, especially for large-scale DNs.

It is vital to accurately estimate the value function in Deep Reinforcement Learning (DRL) such that the agent could execute proper actions instead of suboptimal ones. However, existing actor-critic methods suffer more or less from underestimation bias or overestimation bias, which negatively affect their performance. In this paper, we reveal a simple but effective principle: proper value correction benefits bias alleviation, where we propose the generalized-activated weighting operator that uses any non-decreasing function, namely activation function, as weights for better value estimation. Particularly, we integrate the generalized-activated weighting operator into value estimation and introduce a novel algorithm, Generalized-activated Deep Double Deterministic Policy Gradients (GD3). We theoretically show that GD3 is capable of alleviating the potential estimation bias. We interestingly find that simple activation functions lead to satisfying performance with no additional tricks, and could contribute to faster convergence. Experimental results on numerous challenging continuous control tasks show that GD3 with task-specific activation outperforms the common baseline methods. We also uncover a fact that fine-tuning the polynomial activation function achieves superior results on most of the tasks.

DRL is an AI technique that aims to take appropriate actions to maximise reward in a certain situation (game/simulation/reality). Before further explaining, it is necessary to give some definitions: - Agent: It is the "player" of the game, the entity who's taking actions, he follows a strategy (called policy) to evolve in the environment. His ultimate goal is to maximize his reward. The environment is said to be in a state s at a given time - Policy: It is the strategy which drives the Agent actions, it is designed by a NN. The policy can change as the Agent learns from his experiences - Reward: A metric aiming to determine the performance of the Agent's actions within the environment Now let's take an example to illustrate the mecanisms of DRL: The famous card game of Poker Texas Hold'em (PTH). In PTH, the agents are the players and the environment is the set of rules of PTH (blinds, number of cards, minimum bet, playing order…).

Deep reinforcement learning algorithms can perform poorly in real-world tasks due to the discrepancy between source and target environments. This discrepancy is commonly viewed as the disturbance in transition dynamics. Many existing algorithms learn robust policies by modeling the disturbance and applying it to source environments during training, which usually requires prior knowledge about the disturbance and control of simulators. However, these algorithms can fail in scenarios where the disturbance from target environments is unknown or is intractable to model in simulators. To tackle this problem, we propose a novel model-free actor-critic algorithm -- namely, state-conservative policy optimization (SCPO) -- to learn robust policies without modeling the disturbance in advance. Specifically, SCPO reduces the disturbance in transition dynamics to that in state space and then approximates it by a simple gradient-based regularizer. The appealing features of SCPO include that it is simple to implement and does not require additional knowledge about the disturbance or specially designed simulators. Experiments in several robot control tasks demonstrate that SCPO learns robust policies against the disturbance in transition dynamics.

Quantum computing has a superior advantage in tackling specific problems, such as integer factorization and Simon's problem. For more general tasks in machine learning, by applying variational quantum circuits, more and more quantum algorithms have been proposed recently, especially in supervised learning and unsupervised learning. However, little work has been done in reinforcement learning, arguably more important and challenging. Previous work in quantum reinforcement learning mainly focuses on discrete control tasks where the action space is discrete. In this work, we develop a quantum reinforcement learning algorithm based on soft actor-critic -- one of the state-of-the-art methods for continuous control. Specifically, we use a hybrid quantum-classical policy network consisting of a variational quantum circuit and a classical artificial neural network. Tested in a standard reinforcement learning benchmark, we show that this quantum version of soft actor-critic is comparable with the original soft actor-critic, using much less adjustable parameters. Furthermore, we analyze the effect of different hyper-parameters and policy network architectures, pointing out the importance of architecture design for quantum reinforcement learning.

The potential of reinforcement learning (RL) to deliver aligned and performant agents is partially bottlenecked by the reward engineering problem. One alternative to heuristic trial-and-error is preference-based RL (PbRL), where a reward function is inferred from sparse human feedback. However, prior PbRL methods lack interpretability of the learned reward structure, which hampers the ability to assess robustness and alignment. We propose an online, active preference learning algorithm that constructs reward functions with the intrinsically interpretable, compositional structure of a tree. Using both synthetic and human-provided feedback, we demonstrate sample-efficient learning of tree-structured reward functions in several environments, then harness the enhanced interpretability to explore and debug for alignment.

DB-BERT is a database tuning tool that exploits information gained via natural language analysis of manuals and other relevant text documents. It uses text to identify database system parameters to tune as well as recommended parameter values. DB-BERT applies large, pre-trained language models (specifically, the BERT model) for text analysis. During an initial training phase, it fine-tunes model weights in order to translate natural language hints into recommended settings. At run time, DB-BERT learns to aggregate, adapt, and prioritize hints to achieve optimal performance for a specific database system and benchmark. Both phases are iterative and use reinforcement learning to guide the selection of tuning settings to evaluate (penalizing settings that the database system rejects while rewarding settings that improve performance). In our experiments, we leverage hundreds of text documents about database tuning as input for DB-BERT. We compare DB-BERT against various baselines, considering different benchmarks (TPC-C and TPC-H), metrics (throughput and run time), as well as database systems (Postgres and MySQL). In all cases, DB-BERT finds the best parameter settings among all compared methods. The code of DB-BERT is available online at https://itrummer.github.io/dbbert/.

In this work, we establish heuristics and learning strategies for the autonomous control of a dozer grading an uneven area studded with sand piles. We formalize the problem as a Markov Decision Process, design a simulation which demonstrates agent-environment interactions and finally compare our simulator to a real dozer prototype. We use methods from reinforcement learning, behavior cloning and contrastive learning to train a hybrid policy. Our trained agent, AGPNet, reaches human-level performance and outperforms current state-of-the-art machine learning methods for the autonomous grading task. In addition, our agent is capable of generalizing from random scenarios to unseen real world problems.