Meta-learning has enabled learning statistical models that can be quickly adapted to new prediction tasks. Motivated by use-cases in personalized federated learning, we study the often overlooked aspect of the modern meta-learning algorithms -- their data efficiency. To shed more light on which methods are more efficient, we use techniques from algorithmic stability to derive bounds on the transfer risk that have important practical implications, indicating how much supervision is needed and how it must be allocated for each method to attain the desired level of generalization. Further, we introduce a new simple framework for evaluating meta-learning methods under a limit on the available supervision, conduct an empirical study of MAML, Reptile, and Protonets, and demonstrate the differences in the behavior of these methods on few-shot and federated learning benchmarks. Finally, we propose active meta-learning, which incorporates active data selection into learning-to-learn, leading to better performance of all methods in the limited supervision regime.

Federated learning is typically approached as an optimization problem, where the goal is to minimize a global loss function by distributing computation across client devices that possess local data and specify different parts of the global objective. We present an alternative perspective and formulate federated learning as a posterior inference problem, where the goal is to infer a global posterior distribution by having client devices each infer the posterior of their local data. While exact inference is often intractable, this perspective provides a principled way to search for global optima in federated settings. Further, starting with the analysis of federated quadratic objectives, we develop a computation- and communication-efficient approximate posterior inference algorithm -- federated posterior averaging (FedPA). Our algorithm uses MCMC for approximate inference of local posteriors on the clients and efficiently communicates their statistics to the server, where the latter uses them to refine a global estimate of the posterior mode. Finally, we show that FedPA generalizes federated averaging (FedAvg), can similarly benefit from adaptive optimizers, and yields state-of-the-art results on four realistic and challenging benchmarks, converging faster, to better optima.

Most of the work on interpretable machine learning has focused on designing either inherently interpretable models, which typically trade-off accuracy for interpretability, or post-hoc explanation systems, which lack guarantees about their explanation quality. We propose an alternative to these approaches by directly regularizing a black-box model for interpretability at training time. Our approach explicitly connects three key aspects of interpretable machine learning: (i) the model's innate explainability, (ii) the explanation system used at test time, and (iii) the metrics that measure explanation quality. Our regularization results in substantial improvement in terms of the explanation fidelity and stability metrics across a range of datasets and black-box explanation systems while slightly improving accuracy. Further, if the resulting model is still not sufficiently interpretable, the weight of the regularization term can be adjusted to achieve the desired trade-off between accuracy and interpretability. Finally, we justify theoretically that the benefits of explanation-based regularization generalize to unseen points.

Generalization and reliability of multilingual translation often highly depend on the amount of available parallel data for each language pair of interest. In this paper, we focus on zero-shot generalization---a challenging setup that tests models on translation directions they have not been optimized for at training time. To solve the problem, we (i) reformulate multilingual translation as probabilistic inference, (ii) define the notion of zero-shot consistency and show why standard training often results in models unsuitable for zero-shot tasks, and (iii) introduce a consistent agreement-based training method that encourages the model to produce equivalent translations of parallel sentences in auxiliary languages. We test our multilingual NMT models on multiple public zero-shot translation benchmarks (IWSLT17, UN corpus, Europarl) and show that agreement-based learning often results in 2-3 BLEU zero-shot improvement over strong baselines without any loss in performance on supervised translation directions.

Most work on interpretability in machine learning hasfocused on designing either inherently interpretable models, that typically tradeoff interpretability foraccuracy, or post-hoc explanation systems, that lack guarantees about their explanation quality.We propose an alternative to these approaches by directly regularizing a black-box model for interpretability at training time. Our approach explicitlyconnects three key aspects of interpretable machinelearning: the model's innate explainability, the explanation system used at test time, and the metrics that measure explanation quality. Our regularization results in substantial (up to orders of magnitude) improvement in terms of explanation fidelity and stability metrics across a range of datasets, models, and black-box explanation systems.Remarkably, our regularizers also slightly improve predictive accuracy on average across the nine datasets we consider. Further, we show that the benefits of our novel regularizers on explanation quality provably generalize to unseen test points.

Building agents that can explore their environments intelligently is a challenging open problem. In this paper, we make a step towards understanding how a hierarchical design of the agent's policy can affect its exploration capabilities. First, we design EscapeRoom environments, where the agent must figure out how to navigate to the exit by accomplishing a number of intermediate tasks (\emph{subgoals}), such as finding keys or opening doors. Our environments are procedurally generated and vary in complexity, which can be controlled by the number of subgoals and relationships between them. Next, we propose to measure the complexity of each environment by constructing dependency graphs between the goals and analytically computing \emph{hitting times} of a random walk in the graph. We empirically evaluate Proximal Policy Optimization (PPO) with sparse and shaped rewards, a variation of policy sketches, and a hierarchical version of PPO (called HiPPO) akin to h-DQN. We show that analytically estimated \emph{hitting time} in goal dependency graphs is an informative metric of the environment complexity. We conjecture that the result should hold for environments other than navigation. Finally, we show that solving environments beyond certain level of complexity requires hierarchical approaches.

The score function estimator is widely used for estimating gradients of stochastic objectives in stochastic computation graphs (SCG), eg, in reinforcement learning and meta-learning. While deriving the first-order gradient estimators by differentiating a surrogate loss (SL) objective is computationally and conceptually simple, using the same approach for higher-order derivatives is more challenging. Firstly, analytically deriving and implementing such estimators is laborious and not compliant with automatic differentiation. Secondly, repeatedly applying SL to construct new objectives for each order derivative involves increasingly cumbersome graph manipulations. Lastly, to match the first-order gradient under differentiation, SL treats part of the cost as a fixed sample, which we show leads to missing and wrong terms for estimators of higher-order derivatives. To address all these shortcomings in a unified way, we introduce DiCE, which provides a single objective that can be differentiated repeatedly, generating correct estimators of derivatives of any order in SCGs. Unlike SL, DiCE relies on automatic differentiation for performing the requisite graph manipulations. We verify the correctness of DiCE both through a proof and numerical evaluation of the DiCE derivative estimates. We also use DiCE to propose and evaluate a novel approach for multi-agent learning. Our code is available at https://www.github.com/alshedivat/lola.

Modeling agent behavior is central to understanding the emergence of complex phenomena in multiagent systems. Prior work in agent modeling has largely been task-specific and driven by hand-engineering domain-specific prior knowledge. We propose a general learning framework for modeling agent behavior in any multiagent system using only a handful of interaction data. Our framework casts agent modeling as a representation learning problem. Consequently, we construct a novel objective inspired by imitation learning and agent identification and design an algorithm for unsupervised learning of representations of agent policies. We demonstrate empirically the utility of the proposed framework in (i) a challenging high-dimensional competitive environment for continuous control and (ii) a cooperative environment for communication, on supervised predictive tasks, unsupervised clustering, and policy optimization using deep reinforcement learning.

Multi-agent settings are quickly gathering importance in machine learning. This includes a plethora of recent work on deep multi-agent reinforcement learning, but also can be extended to hierarchical RL, generative adversarial networks and decentralised optimisation. In all these settings the presence of multiple learning agents renders the training problem non-stationary and often leads to unstable training or undesired final results. We present Learning with Opponent-Learning Awareness (LOLA), a method in which each agent shapes the anticipated learning of the other agents in the environment. The LOLA learning rule includes an additional term that accounts for the impact of one agent's policy on the anticipated parameter update of the other agents. Preliminary results show that the encounter of two LOLA agents leads to the emergence of tit-for-tat and therefore cooperation in the iterated prisoners' dilemma, while independent learning does not. In this domain, LOLA also receives higher payouts compared to a naive learner, and is robust against exploitation by higher order gradient-based methods. Applied to repeated matching pennies, LOLA agents converge to the Nash equilibrium. In a round robin tournament we show that LOLA agents can successfully shape the learning of a range of multi-agent learning algorithms from literature, resulting in the highest average returns on the IPD. We also show that the LOLA update rule can be efficiently calculated using an extension of the policy gradient estimator, making the method suitable for model-free RL. This method thus scales to large parameter and input spaces and nonlinear function approximators. We also apply LOLA to a grid world task with an embedded social dilemma using deep recurrent policies and opponent modelling. Again, by explicitly considering the learning of the other agent, LOLA agents learn to cooperate out of self-interest.

We introduce contextual explanation networks (CENs)---a class of models that learn to predict by generating and leveraging intermediate explanations. CENs are deep networks that generate parameters for context-specific probabilistic graphical models which are further used for prediction and play the role of explanations. Contrary to the existing post-hoc model-explanation tools, CENs learn to predict and to explain jointly. Our approach offers two major advantages: (i) for each prediction, valid instance-specific explanations are generated with no computational overhead and (ii) prediction via explanation acts as a regularization and boosts performance in low-resource settings. We prove that local approximations to the decision boundary of our networks are consistent with the generated explanations. Our results on image and text classification and survival analysis tasks demonstrate that CENs are competitive with the state-of-the-art while offering additional insights behind each prediction, valuable for decision support.