Collaborative filtering, a widely-used recommendation technique, predicts a user's preference by aggregating the ratings from similar users. As a result, these measures cannot fully utilize the rating information and are not suitable for real world sparse data. To solve these issues, we propose a novel user distance measure named Preference Mover's Distance (PMD) which makes full use of all ratings made by each user. Our proposed PMD can properly measure the distance between a pair of users even if they have no co-rated items. We show that this measure can be cast as an instance of the Earth Mover's Distance, a well-studied transportation problem for which several highly efficient solvers have been developed. Experimental results show that PMD can help achieve superior recommendation accuracy than state-of-the-art methods, especially when training data is very sparse.

Most existing models of multi-agent reinforcement learning (MARL) adopt centralized training with decentralized execution framework. We demonstrate that the decentralized execution scheme restricts agents' capacity to find a better joint policy in team competition games, where each team of agents share the common rewards and cooperate to compete against other teams. To resolve this problem, we propose Signal Instructed Coordination (SIC), a novel coordination module that can be integrated with most existing models. SIC casts a common signal sampled from a pre-defined distribution to team members, and adopts an information-theoretic regularization to encourage agents to exploit in learning the instruction of centralized signals. Our experiments show that SIC can consistently improve team performance over well-recognized MARL models on matrix games and predator-prey games.

Traditional model-based reinforcement learning approaches learn a model of the environment dynamics without explicitly considering how it will be used by the agent. In the presence of misspecified model classes, this can lead to poor estimates, as some relevant available information is ignored. In this paper, we introduce a novel model-based policy search approach that exploits the knowledge of the current agent policy to learn an approximate transition model, focusing on the portions of the environment that are most relevant for policy improvement. We leverage a weighting scheme, derived from the minimization of the error on the model-based policy gradient estimator, in order to define a suitable objective function that is optimized for learning the approximate transition model. Then, we integrate this procedure into a batch policy improvement algorithm, named Gradient-Aware Model-based Policy Search (GAMPS), which iteratively learns a transition model and uses it, together with the collected trajectories, to compute the new policy parameters. Finally, we empirically validate GAMPS on benchmark domains analyzing and discussing its properties.

Owing to the susceptibility of deep learning systems to adversarial attacks, there has been a great deal of work in developing (both empirically and certifiably) robust classifiers, but the vast majority has defended against single types of attacks. Recent work has looked at defending against multiple attacks, specifically on the MNIST dataset, yet this approach used a relatively complex architecture, claiming that standard adversarial training can not apply because it "overfits" to a particular norm. In this work, we show that it is indeed possible to adversarially train a robust model against a union of norm-bounded attacks, by using a natural generalization of the standard PGD-based procedure for adversarial training to multiple threat models. With this approach, we are able to train standard architectures which are robust against $\ell_\infty$, $\ell_2$, and $\ell_1$ attacks, outperforming past approaches on the MNIST dataset and providing the first CIFAR10 network trained to be simultaneously robust against $(\ell_{\infty}, \ell_{2},\ell_{1})$ threat models, which achieves adversarial accuracy rates of $(47.6\%, 64.8\%, 53.4\%)$ for $(\ell_{\infty}, \ell_{2},\ell_{1})$ perturbations with radius $\epsilon = (0.03,0.5,12)$.

Many real-world problems can be reduced to combinatorial optimization on a graph, where the subset or ordering of vertices that maximize some objective function must be found. With such tasks often NP-hard and analytically intractable, reinforcement learning (RL) has shown promise as a framework with which efficient heuristic methods to tackle these problems can be learned. Previous works construct the solution subset incrementally, adding one element at a time, however, the irreversible nature of this approach prevents the agent from revising its earlier decisions, which may be necessary given the complexity of the optimization task. We instead propose that the agent should seek to continuously improve the solution by learning to explore at test time. Our approach of exploratory combinatorial optimization (ECO-DQN) is, in principle, applicable to any combinatorial problem that can be defined on a graph. Experimentally, we show our method to produce state-of-the-art RL performance on the Maximum Cut problem. Moreover, because ECO-DQN can start from any arbitrary configuration, it can be combined with other search methods to further improve performance, which we demonstrate using a simple random search.

We study the problem of identifying the policy space of a learning agent, having access to a set of demonstrations generated by its optimal policy. We introduce an approach based on statistical testing to identify the set of policy parameters the agent can control, within a larger parametric policy space. After presenting two identification rules (combinatorial and simplified), applicable under different assumptions on the policy space, we provide a probabilistic analysis of the simplified one in the case of linear policies belonging to the exponential family. To improve the performance of our identification rules, we frame the problem in the recently introduced framework of the Configurable Markov Decision Processes, exploiting the opportunity of configuring the environment to induce the agent revealing which parameters it can control. Finally, we provide an empirical evaluation, on both discrete and continuous domains, to prove the effectiveness of our identification rules.

Low Back Pain (LBP) is a common medical condition that deprives many individuals worldwide of their normal routine activities. In the absence of external biomarkers, diagnosis of LBP is quite challenging. It requires dealing with several clinical variables, which have no precisely quantified values. Aiming at the development of a fuzzy medical expert system for LBP management, this research proposes an attractive lattice-based knowledge representation scheme for handling imprecision in knowledge, offering a suitable design methodology for a fuzzy knowledge base and a fuzzy inference system. The fuzzy knowledge base is constructed in modular fashion, with each module capturing interrelated medical knowledge about the relevant clinical history, clinical examinations and laboratory investigation results. This approach in design ensures optimality, consistency and preciseness in the knowledge base and scalability. The fuzzy inference system, which uses the Mamdani method, adopts the triangular membership function for fuzzification and the Centroid of Area technique for defuzzification. A prototype of this system has been built using the knowledge extracted from the domain expert physicians. The inference of the system against a few available patient records at the ESI Hospital, Sealdah has been checked. It was found to be acceptable by the verifying medical experts.

The paper presents a new method for approximating Strong Stackelberg Equilibrium in general-sum sequential games with imperfect information and perfect recall. The proposed approach is generic as it does not rely on any specific properties of a particular game model. The method is based on iterative interleaving of the two following phases: (1) guided Monte Carlo Tree Search sampling of the Follower's strategy space and (2) building the Leader's behavior strategy tree for which the sampled Follower's strategy is an optimal response. The above solution scheme is evaluated with respect to expected Leader's utility and time requirements on three sets of interception games with variable characteristics, played on graphs. A comparison with three state-of-the-art MILP/LP-based methods shows that in vast majority of test cases proposed simulation-based approach leads to optimal Leader's strategies, while excelling the competitive methods in terms of better time scalability and lower memory requirements.

We explore fixed-horizon temporal difference (TD) methods, reinforcement learning algorithms for a new kind of value function that predicts the sum of rewards over a $\textit{fixed}$ number of future time steps. To learn the value function for horizon $h$, these algorithms bootstrap from the value function for horizon $h-1$, or some shorter horizon. Because no value function bootstraps from itself, fixed-horizon methods are immune to the stability problems that plague other off-policy TD methods using function approximation (also known as "the deadly triad"). Although fixed-horizon methods require the storage of additional value functions, this gives the agent additional predictive power, while the added complexity can be substantially reduced via parallel updates, shared weights, and $n$-step bootstrapping. We show how to use fixed-horizon value functions to solve reinforcement learning problems competitively with methods such as Q-learning that learn conventional value functions. We also prove convergence of fixed-horizon temporal difference methods with linear and general function approximation. Taken together, our results establish fixed-horizon TD methods as a viable new way of avoiding the stability problems of the deadly triad.

In natural language understanding components, detecting out-of-domain (OOD) inputs is important for dialogue systems since wrongly accepting these OOD utterances that are not currently supported may lead to catastrophic failures of the entire system. Entropy regularization is an effective solution to avoid such failures, however, its computation heavily depends on OOD data, which are expensive to collect. In this paper, we propose a novel text generation model to produce high-quality OOD samples and thereby improve the performance of OOD detection. The proposed model can also utilize a set of unlabeled data to improve the effectiveness of these generated OOD samples. Experiments show that our method can effectively improve the OOD detection performance of a NLU module. 1 Introduction Natural Language Understanding (NLU) in dialog systems, particularly including task-oriented dialog systems and intelligent personal assistants, is vital for understanding users' inputs and making effective interactions. NLU maps text inputs to structured user intents, and decides the downstream processing pipelines of a dialog system, thereby becoming a precursor for the success of such systems. Recently, various deep neural network (DNN) based NLU models have been proposed and applied in real-world applications (Kim et al., 2018; Sarikaya, 2017; Y oo et al., 2018). Most existing DNN based NLU modules are built by following a closed-world assumption (Fei and Liu, 2016), i.e, the data used in the training and test phrase are drawn from the same distribution. However, such an assumption is commonly violated in practical systems that are deployed in a dynamic or open environment. Specifically, practical NLU systems often encounter o ut-o f-d omain (OOD) inputs that are not supported by the system and thus not observed in the training data.