We show how an ensemble of $Q^*$-functions can be leveraged for more effective exploration in deep reinforcement learning. We build on well established algorithms from the bandit setting, and adapt them to the $Q$-learning setting. We propose an exploration strategy based on upper-confidence bounds (UCB). Our experiments show significant gains on the Atari benchmark.

We are currently seeking a hands-on Machine Learning Scientist (Distributed Systems, Tensorflow) for our new research-led startup, focussing on the application of artificial intelligence in the real world; particularly smart city simulations and bots. We're looking for a hardcore Machine Learning Scientist/Engineer who thrives wants to work with the latest technology in multi-agent learning algorithms, Gaussian process and reinforcement learning. As a Machine Learning Scientist/Engineer, you will be a core member of the machine learning team; working closely with the Machine Learning researchers, transforming their algorithmic research into highly innovative products which will be attractive and accessible to the world. Key Skills: Machine Learning Engineer/ML Scientist, Tensorflow, C, C, Java, Python, C#, Distributed Algorithms. Distributed systems, BSc, MSc, MPhil, PhD, Post-Doc, Research, R&D, startup, Multithreading.

Reinforcement Learning (RL) can model complex behavior policies for goal-directed sequential decision making tasks. A hallmark of RL algorithms is Temporal Difference (TD) learning: value function for the current state is moved towards a bootstrapped target that is estimated using next state's value function. $\lambda$-returns generalize beyond 1-step returns and strike a balance between Monte Carlo and TD learning methods. While lambda-returns have been extensively studied in RL, they haven't been explored a lot in Deep RL. This paper's first contribution is an exhaustive benchmarking of lambda-returns. Although mathematically tractable, the use of exponentially decaying weighting of n-step returns based targets in lambda-returns is a rather ad-hoc design choice. Our second major contribution is that we propose a generalization of lambda-returns called Confidence-based Autodidactic Returns (CAR), wherein the RL agent learns the weighting of the n-step returns in an end-to-end manner. This allows the agent to learn to decide how much it wants to weigh the n-step returns based targets. In contrast, lambda-returns restrict RL agents to use an exponentially decaying weighting scheme. Autodidactic returns can be used for improving any RL algorithm which uses TD learning. We empirically demonstrate that using sophisticated weighted mixtures of multi-step returns (like CAR and lambda-returns) considerably outperforms the use of n-step returns. We perform our experiments on the Asynchronous Advantage Actor Critic (A3C) algorithm in the Atari 2600 domain.

Temporal-difference (TD) learning is an important field in reinforcement learning. Sarsa and Q-Learning are among the most used TD algorithms. The Q($\sigma$) algorithm (Sutton and Barto (2017)) unifies both. This paper extends the Q($\sigma$) algorithm to an online multi-step algorithm Q($\sigma, \lambda$) using eligibility traces and introduces Double Q($\sigma$) as the extension of Q($\sigma$) to double learning. Experiments suggest that the new Q($\sigma, \lambda$) algorithm can outperform the classical TD control methods Sarsa($\lambda$), Q($\lambda$) and Q($\sigma$).

We present MILABOT: a deep reinforcement learning chatbot developed by the Montreal Institute for Learning Algorithms (MILA) for the Amazon Alexa Prize competition. MILABOT is capable of conversing with humans on popular small talk topics through both speech and text. The system consists of an ensemble of natural language generation and retrieval models, including template-based models, bag-of-words models, sequence-to-sequence neural network and latent variable neural network models. By applying reinforcement learning to crowdsourced data and real-world user interactions, the system has been trained to select an appropriate response from the models in its ensemble. The system has been evaluated through A/B testing with real-world users, where it performed significantly better than many competing systems. Due to its machine learning architecture, the system is likely to improve with additional data.

Reinforcement learning is an increasingly popular machine learning technique that is particularly well suited for addressing problems within dynamic and adaptive environments. When paired with simulations, reinforcement learning is a powerful tool for training AI models that can help increase automation or optimize operational efficiency of sophisticated systems such as robotics, manufacturing, and supply chain logistics. However, moving from the games commonly used to demonstrate these techniques into real-world applications isn't always straightforward. Structuring solutions to move beyond purely data-driven training introduces all sorts of new complexity, requiring you to consider things like how to use simulations to target your learning objectives, what kinds of simulations are applicable, how to deal with long-running simulations, how to incorporate ongoing training refinement once deployed, how to account for scaling and performance, and ultimately how to bridge from simulation to the real world. I was recently able to talk about how to effectively leverage reinforcement learning in real-world use cases at the O'Reilly AI conference in San Francisco.

Conventional reinforcement learning methods for Markov decision processes rely on weakly-guided, stochastic searches to drive the learning process. It can therefore be difficult to predict what agent behaviors might emerge. In this paper, we consider an information-theoretic cost function for performing constrained stochastic searches that promote the formation of risk-averse to risk-favoring behaviors. This cost function is the value of information, which provides the optimal trade-off between the expected return of a policy and the policy's complexity; policy complexity is measured by number of bits and controlled by a single hyperparameter on the cost function. As the policy complexity is reduced, the agents will increasingly eschew risky actions. This reduces the potential for high accrued rewards. As the policy complexity increases, the agents will take actions, regardless of the risk, that can raise the long-term rewards. The obtainable reward depends on a single, tunable hyperparameter that regulates the degree of policy complexity. We evaluate the performance of value-of-information-based policies on a stochastic version of Ms. Pac-Man. A major component of this paper is the demonstration that ranges of policy complexity values yield different game-play styles and explaining why this occurs. We also show that our reinforcement-learning search mechanism is more efficient than the others we utilize. This result implies that the value of information theory is appropriate for framing the exploitation-exploration trade-off in reinforcement learning.

We introduce a novel repeated Inverse Reinforcement Learning problem: the agent has to act on behalf of a human in a sequence of tasks and wishes to minimize the number of tasks that it surprises the human by acting suboptimally with respect to how the human would have acted. Each time the human is surprised, the agent is provided a demonstration of the desired behavior by the human. We formalize this problem, including how the sequence of tasks is chosen, in a few different ways and provide some foundational results.

Accumulating evidence suggest that human behavior in trial-and-error learning tasks based on decisions between discrete actions may involve a combination of reinforcement learning (RL) and working-memory (WM). While the understanding of brain activity at stake in this type of tasks often involve the comparison with non-human primate neurophysiological results, it is not clear whether monkeys use similar combined RL and WM processes to solve these tasks. Here we analyzed the behavior of five monkeys with computational models combining RL and WM. Our model-based analysis approach enables to not only fit trial-by-trial choices but also transient slowdowns in reaction times, indicative of WM use. We found that the behavior of the five monkeys was better explained in terms of a combination of RL and WM despite inter-individual differences. The same coordination dynamics we used in a previous study in humans best explained the behavior of some monkeys while the behavior of others showed the opposite pattern, revealing a possible different dynamics of WM process. We further analyzed different variants of the tested models to open a discussion on how the long pretraining in these tasks may have favored particular coordination dynamics between RL and WM. This points towards either inter-species differences or protocol differences which could be further tested in humans.

Deep reinforcement learning (DRL) methods such as the Deep Q-Network (DQN) have achieved state-of-the-art results in a variety of challenging, high-dimensional domains. This success is mainly attributed to the power of deep neural networks to learn rich domain representations for approximating the value function or policy. Batch reinforcement learning methods with linear representations, on the other hand, are more stable and require less hyper parameter tuning. Yet, substantial feature engineering is necessary to achieve good results. In this work we propose a hybrid approach -- the Least Squares Deep Q-Network (LS-DQN), which combines rich feature representations learned by a DRL algorithm with the stability of a linear least squares method. We do this by periodically re-training the last hidden layer of a DRL network with a batch least squares update. Key to our approach is a Bayesian regularization term for the least squares update, which prevents over-fitting to the more recent data. We tested LS-DQN on five Atari games and demonstrate significant improvement over vanilla DQN and Double-DQN. We also investigated the reasons for the superior performance of our method. Interestingly, we found that the performance improvement can be attributed to the large batch size used by the LS method when optimizing the last layer.