In this work, we propose a novel cross Q-learning algorithm, aim at alleviating the well-known overestimation problem in value-based reinforcement learning methods, particularly in the deep Q-networks where the overestimation is exaggerated by function approximation errors. Our algorithm builds on double Q-learning, by maintaining a set of parallel models and estimate the Q-value based on a randomly selected network, which leads to reduced overestimation bias as well as the variance. We provide empirical evidence on the advantages of our method by evaluating on some benchmark environment, the experimental results demonstrate significant improvement of performance in reducing the overestimation bias and stabilizing the training, further leading to better derived policies.

Imtiaz Adam, Twitter @Deeplearn007 Updated a few sections in Sep 2020 Artificial Intelligence (AI) is increasingly affecting the world around us. It is increasingly making an impact in retail, financial services, along with other sectors of the economy.

The "Curly" curling robots are capturing hearts around the world. A product of Korea University in Seoul and the Berlin Institute of Technology, the deep reinforcement learning powered bots slide stones along ice in a winter sport that dates to the 16th century. As much as their human-expert-bettering accuracy or technology impresses, a big part of the Curly appeal is how we see the little machines in the physical space: the determined manner in which the thrower advances in the arena, smartly raising its head-like cameras to survey the shiny white curling sheet, gently cradling and rotating a rock to begin delivery, releasing deftly at the hog line as a skip watches from the backline, with our hopes. Artificial intelligence (AI) today delivers everything from soup recipes to stock predictions, but most tech works out-of-sight. More visible are the physical robots of various shapes, sizes and functions that embody the latest AI technologies. These robots have generally been helpful, and now they are also becoming a more entertaining and enjoyable part of our lives.

Episodic reinforcement learning and contextual bandits are two widely studied sequential decision-making problems. Episodic reinforcement learning generalizes contextual bandits and is often perceived to be more difficult due to long planning horizon and unknown state-dependent transitions. The current paper shows that the long planning horizon and the unknown state-dependent transitions (at most) pose little additional difficulty on sample complexity. We consider the episodic reinforcement learning with $S$ states, $A$ actions, planning horizon $H$, total reward bounded by $1$, and the agent plays for $K$ episodes. We propose a new algorithm, \textbf{M}onotonic \textbf{V}alue \textbf{P}ropagation (MVP), which relies on a new Bernstein-type bonus. The new bonus only requires tweaking the \emph{constants} to ensure optimism and thus is significantly simpler than existing bonus constructions. We show MVP enjoys an $O\left(\left(\sqrt{SAK} + S^2A\right) \text{poly}\log \left(SAHK\right)\right)$ regret, approaching the $\Omega\left(\sqrt{SAK}\right)$ lower bound of \emph{contextual bandits}. Notably, this result 1) \emph{exponentially} improves the state-of-the-art polynomial-time algorithms by Dann et al. [2019], Zanette et al. [2019] and Zhang et al. [2020] in terms of the dependency on $H$, and 2) \emph{exponentially} improves the running time in [Wang et al. 2020] and significantly improves the dependency on $S$, $A$ and $K$ in sample complexity.

Model-free algorithms such as Q-learning (Watkins and Dayan, 1992) play a central role in recent breakthroughs of reinforcement learning (RL) (Mnih et al., 2015). In contrast to model-based algorithms that decouple model estimation and planning, model-free algorithms attempt to directly interact with the environment -- in the form of a policy that selects actions based on perceived states of the environment -- from the collected data samples, without modeling the environment explicitly. Therefore, model-free algorithms are able to process data in an online fashion and are often memory-efficient. Understanding and improving the sample efficiency of model-free algorithms lie at the core of recent research activity (Dulac-Arnold et al., 2019), whose importance is particularly evident for the class of RL applications in which data collection is costly and time-consuming (such as clinical trials, online advertisements, and so on). The current paper concentrates on Q-learning -- an off-policy model-free algorithm that seeks to learn the optimal action-value function by observing what happens under a behavior policy. The off-policy feature makes it appealing in various RL applications where it is infeasible to change the policy under evaluation on the fly. There are two basic update models in Q-learning. The first one is termed a synchronous setting, which hypothesizes on the existence of a simulator (or a generative model); at each time, the simulator generates an independent sample for every state-action pair, and the estimates are updated simultaneously across all state-action pairs. The second model concerns an asynchronous setting, where only a single sample trajectory following a behavior policy is accessible; at each time, the algorithm updates its estimate of a single state-action pair using one state transition from the trajectory.

Experience replay (ER) improves the data efficiency of off-policy reinforcement learning (RL) algorithms by allowing an agent to store and reuse its past experiences in a replay buffer. While many techniques have been proposed to enhance ER by biasing how experiences are sampled from the buffer, thus far they have not considered strategies for refreshing experiences inside the buffer. In this work, we introduce Lucid Dreaming for Experience Replay (LiDER), a conceptually new framework that allows replay experiences to be refreshed by leveraging the agent's current policy. LiDER 1) moves an agent back to a past state; 2) lets the agent try following its current policy to execute different actions---as if the agent were "dreaming" about the past, but is aware of the situation and can control the dream to encounter new experiences; and 3) stores and reuses the new experience if it turned out better than what the agent previously experienced, i.e., to refresh its memories. LiDER is designed to be easily incorporated into off-policy, multi-worker RL algorithms that use ER; we present in this work a case study of applying LiDER to an actor-critic based algorithm. Results show LiDER consistently improves performance over the baseline in four Atari 2600 games. Our open-source implementation of LiDER and the data used to generate all plots in this paper are available at github.com/duyunshu/lucid-dreaming-for-exp-replay.

The main drawback to reinforcement learning is that it can't be used in some real-life applications. That's because in the process of training themselves, computers initially try just about anything and everything before eventually stumbling on the right path. This initial trial-and-error phase can be problematic for certain applications, such as climate-control systems where abrupt swings in temperature wouldn't be tolerated. The CSEM engineers have developed an approach that overcomes this problem. They showed that computers can first be trained on extremely simplified theoretical models before being set to learn on real-life systems.

Multicasting in wireless systems is a natural way to exploit the redundancy in user requests in a Content Centric Network. Power control and optimal scheduling can significantly improve the wireless multicast network's performance under fading. However, the model based approaches for power control and scheduling studied earlier are not scalable to large state space or changing system dynamics. In this paper, we use deep reinforcement learning where we use function approximation of the Q-function via a deep neural network to obtain a power control policy that matches the optimal policy for a small network. We show that power control policy can be learnt for reasonably large systems via this approach. Further we use multi-timescale stochastic optimization to maintain the average power constraint. We demonstrate that a slight modification of the learning algorithm allows tracking of time varying system statistics. Finally, we extend the multi-timescale approach to simultaneously learn the optimal queueing strategy along with power control. We demonstrate scalability, tracking and cross layer optimization capabilities of our algorithms via simulations. The proposed multi-timescale approach can be used in general large state space dynamical systems with multiple objectives and constraints, and may be of independent interest.

This paper studies the problem of learning the correlation structure of a set of intervention functions defined on the directed acyclic graph (DAG) of a causal model. This is useful when we are interested in jointly learning the causal effects of interventions on different subsets of variables in a DAG, which is common in field such as healthcare or operations research. We propose the first multi-task causal Gaussian process (GP) model, which we call DAG-GP, that allows for information sharing across continuous interventions and across experiments on different variables. DAG-GP accommodates different assumptions in terms of data availability and captures the correlation between functions lying in input spaces of different dimensionality via a well-defined integral operator. We give theoretical results detailing when and how the DAG-GP model can be formulated depending on the DAG. We test both the quality of its predictions and its calibrated uncertainties. Compared to single-task models, DAG-GP achieves the best fitting performance in a variety of real and synthetic settings. In addition, it helps to select optimal interventions faster than competing approaches when used within sequential decision making frameworks, like active learning or Bayesian optimization.

Recent advancements in quantum computing have driven the scientific community's quest to solve a certain class of complex problems for which quantum computers would be better suited than traditional supercomputers. To improve the efficiency with which quantum computers can solve these problems, scientists are investigating the use of artificial intelligence approaches. In a new study, scientists at the U.S. Department of Energy's (DOE) Argonne National Laboratory have developed a new algorithm based on reinforcement learning to find the optimal parameters for the Quantum Approximate Optimization Algorithm (QAOA), which allows a quantum computer to solve certain combinatorial problems such as those that arise in materials design, chemistry and wireless communications. "Combinatorial optimization problems are those for which the solution space gets exponentially larger as you expand the number of decision variables," said Argonne computer scientist Prasanna Balaprakash. "In one traditional example, you can find the shortest route for a salesman who needs to visit a few cities once by enumerating all possible routes, but given a couple thousand cities, the number of possible routes far exceeds the number of stars in the universe; even the fastest supercomputers cannot find the shortest route in a reasonable time."