We introduce two tactics to attack agents trained by deep reinforcement learning algorithms using adversarial examples, namely the strategically-timed attack and the enchanting attack. In the strategically-timed attack, the adversary aims at minimizing the agent's reward by only attacking the agent at a small subset of time steps in an episode. Limiting the attack activity to this subset helps prevent detection of the attack by the agent. We propose a novel method to determine when an adversarial example should be crafted and applied. In the enchanting attack, the adversary aims at luring the agent to a designated target state. This is achieved by combining a generative model and a planning algorithm: while the generative model predicts the future states, the planning algorithm generates a preferred sequence of actions for luring the agent. A sequence of adversarial examples is then crafted to lure the agent to take the preferred sequence of actions. We apply the two tactics to the agents trained by the state-of-the-art deep reinforcement learning algorithm including DQN and A3C. In 5 Atari games, our strategically timed attack reduces as much reward as the uniform attack (i.e., attacking at every time step) does by attacking the agent 4 times less often. Our enchanting attack lures the agent toward designated target states with a more than 70% success rate. Videos are available at http://yenchenlin.me/adversarial_attack_RL/

We study incentivized exploration for the multi-armed bandit (MAB) problem where the players receive compensation for exploring arms other than the greedy choice and may provide biased feedback on reward. We seek to understand the impact of this drifted reward feedback by analyzing the performance of three instantiations of the incentivized MAB algorithm: UCB, $\varepsilon$-Greedy, and Thompson Sampling. Our results show that they all achieve $\mathcal{O}(\log T)$ regret and compensation under the drifted reward, and are therefore effective in incentivizing exploration. Numerical examples are provided to complement the theoretical analysis.

Reinforcement learning is effective in optimizing policies for recommender systems. Current solutions mostly focus on model-free approaches, which require frequent interactions with a real environment, and thus are expensive in model learning. Offline evaluation methods, such as importance sampling, can alleviate such limitations, but usually request a large amount of logged data and do not work well when the action space is large. In this work, we propose a model-based reinforcement learning solution which models the user-agent interaction for offline policy learning via a generative adversarial network. To reduce bias in the learnt policy, we use the discriminator to evaluate the quality of generated sequences and rescale the generated rewards. Our theoretical analysis and empirical evaluations demonstrate the effectiveness of our solution in identifying patterns from given offline data and learning policies based on the offline and generated data.

The goal of policy gradient approaches is to find a policy in a given class of policies which maximizes the expected return. Given a differentiable model of the policy, we want to apply a gradient-ascent technique to reach a local optimum. We mainly use gradient ascent, because it is theoretically well researched. The main issue is that the policy gradient with respect to the expected return is not available, thus we need to estimate it. As policy gradient algorithms also tend to require on-policy data for the gradient estimate, their biggest weakness is sample efficiency. For this reason, most research is focused on finding algorithms with improved sample efficiency. This paper provides a formal introduction to policy gradient that shows the development of policy gradient approaches, and should enable the reader to follow current research on the topic.

Requests come in a variety of size and constraints, from the urgent monitoring of small areas to large area coverage. In this work we are particularly interested in the latter case, with requests covering whole countries or even continents. Depending on weather conditions, such requests may take several months to complete, even with multiple satellites. In order to shorten the time required to fulfill requests, the mission orchestrator shall schedule acquisitions with both a short and a long-term strategy. Determining a strategy robust to an uncertain environment is a complex task, this is why current solutions mainly consist of heuristics configured by human-experts. This paper demonstrates that Reinforcement Learning (RL) might be well-suited for such a challenge. RL has proven to be of great value since these algorithms have mastered several games such as Pong on Atari 2600 (Mnih et al. 2013), Go with AlphaGo (Silver et al. 2017) and more recently Starcraft (Arulkumaran, Cully, and Togelius 2019).c

Under dynamic traffic, service function chain (SFC) migration is considered as an effective way to improve resource utilization. However, the lack of future network information leads to non-optimal solutions, which motivates us to study reinforcement learning based SFC migration from a long-term perspective. In this paper, we formulate the SFC migration problem as a minimization problem with the objective of total network operation cost under constraints of users' quality of service. We firstly design a deep Q-network based algorithm to solve single SFC migration problem, which can adjust migration strategy online without knowing future information. Further, a novel multi-agent cooperative framework, called MSDF, is proposed to address the challenge of considering multiple SFC migration on the basis of single SFC migration. MSDF reduces the complexity thus accelerates the convergence speed, especially in large scale networks. Experimental results demonstrate that MSDF outperforms typical heuristic algorithms under various scenarios.

Abstract-- Logic synthesis requires extensive tuning of the synthesis optimization flow where the quality of results (QoR) depends on the sequence of optimizations used. Efficient design space exploration is challenging due to the exponential number of possible optimization permutations. Therefore, automating the optimization process is necessary. In this work, we propose a novel reinforcement learning-based methodology that navigates the optimization space without human intervention. We demonstrate the training of an Advantage Actor Critic (A2C) agent that seeks to minimize area subject to a timing constraint. Using the proposed methodology, designs can be optimized autonomously with no-humans in-loop. Evaluation on the comprehensive EPFL benchmark suite shows that the agent outperforms existing exploration methodologies and improves QoRs by an average of 13%. Logic synthesis transforms a high-level description of a design into an optimized gate-level representation. Modern logic synthesis tools represent a given design as an And-Inverter Graph (AIG), which encodes representative characteristics for optimizing Boolean functions.

Massive IoT including the large number of resource-constrained IoT devices has gained great attention. IoT devices generate enormous traffic, which causes network congestion. To manage network congestion, multi-channel-based algorithms are proposed. However, most of the existing multi-channel algorithms require strict synchronization, an extra overhead for negotiating channel assignment, which poses significant challenges to resource-constrained IoT devices. In this paper, a distributed channel selection algorithm utilizing the tug-of-war (TOW) dynamics is proposed for improving successful frame delivery of the whole network by letting IoT devices always select suitable channels for communication adaptively. The proposed TOW dynamics-based channel selection algorithm has a simple reinforcement learning procedure that only needs to receive the acknowledgment (ACK) frame for the learning procedure, while simply requiring minimal memory and computation capability.

We propose a conceptually simple and lightweight framework for deep reinforcement learning that uses asynchronous gradient descent for optimization of deep neural network controllers. We present asynchronous variants of four standard reinforcement learning algorithms and show that parallel actor-learners have a stabilizing effect on training allowing all four methods to successfully train neural network controllers.

The use of an environment means that there is no fixed training dataset, rather a goal or set of goals that an agent is required to achieve, actions they may perform, and feedback about performance toward the goal. Some machine learning algorithms do not just experience a fixed dataset. For example, reinforcement learning algorithms interact with an environment, so there is a feedback loop between the learning system and its experiences.