UAI 2021 Papers

Meta-learning learns across historical tasks with the goal to discover a representation from which it is easy to adapt to unseen tasks. Episodic meta-learning attempts to simulate a realistic setting by generating a set of small artificial tasks from a larger set of training tasks for meta-training and proceeds in a similar fashion for meta-testing. However, this (meta-)learning paradigm has recently been shown to be brittle, suggesting that the inductive bias encoded in the learned representations is inadequate. In this work we propose to compose episodes to robustify meta-learning in the few-shot setting in order to learn more efficiently and to generalize better to new tasks. We make use of active learning scoring rules to select the data to be included in the episodes. We assume that the meta-learner is given new tasks at random, but the data associated to the tasks can be selected from a larger pool of unlabeled data, and investigate where active learning can boost the performance of episodic meta-learning. We show that instead of selecting samples at random, it is better to select samples in an active manner especially in settings with out-of-distribution and class-imbalanced tasks. We evaluate our method with Prototypical Networks, foMAML and protoMAML, reporting significant improvements on public benchmarks.

As the representations output by Graph Neural Networks (GNNs) are increasingly employed in real-world applications, it becomes important to ensure that these representations are fair and stable. In this work, we establish a key connection between counterfactual fairness and stability and leverage it to propose a novel framework, NIFTY (uNIfying Fairness and stabiliTY), which can be used with any GNN to learn fair and stable representations. We introduce a novel objective function that simultaneously accounts for fairness and stability and develop a layer-wise weight normalization using the Lipschitz constant to enhance neural message passing in GNNs. In doing so, we enforce fairness and stability both in the objective function as well as in the GNN architecture. Further, we show theoretically that our layer-wise weight normalization promotes counterfactual fairness and stability in the resulting representations. We introduce three new graph datasets comprising of high-stakes decisions in criminal justice and financial lending domains. Extensive experimentation with the above datasets demonstrates the efficacy of our framework.

In most of the transfer learning approaches to reinforcement learning (RL) the distribution over the tasks is assumed to be stationary. Therefore, the target and source tasks are i.i.d. samples of the same distribution. Unfortunately, this assumption rarely holds in real-world conditions, e.g., due to seasonality or periodicity, evolution in the environment or faults in the sensors/actuators. In the context of this work, we consider the problem of transferring value functions through a variational method when the distribution that generates the tasks is time-variant, proposing a solution that leverages this temporal structure inherent in the task generating process. Furthermore, by means of a finite-sample analysis, the previously mentioned solution is theoretically compared to its time-invariant version. Finally, the experimental evaluation of the proposed technique is carried out on the lake Como water system representing a real-world scenario and on three different RL environments with three distinct temporal dynamics.

This paper studies the relationship between generalization and privacy preservation of machine learning in two steps. We first establish an alignment between the two facets for any learning algorithm. We prove that $(\varepsilon, \delta)$-differential privacy implies an on-average generalization bound for a multi-sample-set learning algorithm, which further leads to a high-probability bound for any learning algorithm. We then investigate how the iterative nature shared by most learning algorithms influences privacy preservation and further generalization. Three composition theorems are proved to approximate the differential privacy of an iterative algorithm through the differential privacy of its every iteration. Integrating the above two steps, we eventually deliver generalization bounds for iterative learning algorithms. Our results are strictly tighter than the existing works. Particularly, our generalization bounds do not rely on the model size which is prohibitively large in deep learning. Experiments of MLP, VGG, and ResNet on MNIST, CIFAR-10, and CIFAR-100 are in full agreement with our theory. The theory applies to a wide spectrum of learning algorithms. In this paper, it is applied to the Gaussian mechanism as an example.

Reinforcement learning problems with multiple agents pose the challenge of efficiently adapting to nonstationary dynamics arising from other agents’ strategic behavior. Although several algorithms exist for these problems with promising empirical results, regret analysis and efficient use of other-agent models in general-sum games is very limited. We propose an algorithm (TSMG) for general-sum Markov games against agents that switch between several stationary policies, combining change detection with Thompson sampling to learn parametric models of these policies. Under standard assumptions for parametric Markov decision process (MDP) learning, the expected regret of TSMG in the worst case over policy parameters and switch schedules is near-optimal in time and number of switches, up to logarithmic factors. Our experiments on simulated games show that TSMG can outperform standard Thompson sampling and a version of Thompson sampling with a static reset schedule, despite the violation of an assumption that the MDPs induced by the other player are ergodic.

Deep kernel learning and related techniques promise to combine the representational power of neural networks with the reliable uncertainty estimates of Gaussian processes. One crucial aspect of these models is an expectation that, because they are treated as Gaussian process models optimized using the marginal likelihood, they are protected from overfitting. However, we identify pathological behavior, including overfitting, on a simple toy example. We explore this pathology, explaining its origins and considering how it applies to real datasets. Through careful experimentation on UCI datasets, CIFAR-10, and the UTKFace dataset, we find that the overfitting from overparameterized deep kernel learning, in which the model is “somewhat Bayesian”, can in certain scenarios be worse than that from not being Bayesian at all. However, we find that a fully Bayesian treatment of deep kernel learning can rectify this overfitting and obtain the desired performance improvements over standard neural networks and Gaussian processes.

Variational inference has had great success in scaling approximate Bayesian inference to big data by exploiting mini-batch training. To date, however, this strategy has been most applicable to models of independent data. We propose an extension to state space models of time series data based on a novel generative model for latent temporal states: the neural moving average model. This permits a subsequence to be sampled without drawing from the entire distribution, enabling training iterations to use mini-batches of the time series at low computational cost. We illustrate our method on autoregressive, Lotka-Volterra, FitzHugh-Nagumo and stochastic volatility models, achieving accurate parameter estimation in a short time.

Adversarial training has shown its ability in producing models that are robust to perturbations on the input data, but usually at the expense of a decrease in the standard accuracy. To mitigate this issue, it is commonly believed that more training data will eventually help such adversarially robust models generalize better on the benign/unperturbed test data. In this paper, however, we challenge this conventional belief and show that more training data can hurt the generalization of adversarially robust models in classification problems. We first investigate the Gaussian mixture classification with a linear loss and identify three regimes based on the strength of the adversary. In the weak adversary regime, more data improves the generalization of adversarially robust models. In the medium adversary regime, with more training data, the generalization loss exhibits a double descent curve, which implies the existence of an intermediate stage where more training data hurts the generalization. In the strong adversary regime, more data almost immediately causes the generalization error to increase. Then we analyze a two-dimensional classification problem with a 0-1 loss. We prove that more data always hurts generalization of adversarially trained models with large perturbations. Empirical studies confirm our theoretical results.

This paper studies the complexity for finding approximate stationary points of nonconvex-strongly-concave (NC-SC) smooth minimax problems, in both general and averaged smooth finite-sum settings. We establish nontrivial lower complexity bounds for the two settings, respectively. Our result reveals substantial gaps between these limits and best-known upper bounds in the literature. To close these gaps, we introduce a generic acceleration scheme that deploys existing gradient-based methods to solve a sequence of crafted strongly-convex-strongly-concave subproblems. In the general setting, the complexity of our proposed algorithm nearly matches the lower bound; in particular, it removes an additional poly-logarithmic dependence on accuracy present in previous works. In the averaged smooth finite-sum setting, our proposed algorithm improves over previous algorithms by providing a nearly-tight dependence on the condition number.

Several algorithms for finding the best arm in the dueling bandits setting assume the existence of a Condorcet winner (CW), that is, an arm that uniformly dominates all other arms. Yet, by simply relying on this assumption but not verifying it, such algorithms may produce doubtful results in cases where it actually fails to hold. Even worse, the problem may not be noticed, and an alleged CW still be produced. In this paper, we therefore address the problem as a ”testification” task, by which we mean a combination of testing and identification: The online identification of the CW is combined with the statistical testing of the CW assumption. Thus, instead of returning a supposed CW at some point, the learner has the possibility to stop sampling and refuse an answer in case it feels confident that the CW assumption is violated. Analyzing the testification problem formally, we derive lower bounds on the expected sample complexity of any online algorithm solving it. Moreover, a concrete algorithm is proposed, which achieves the optimal sample complexity up to logarithmic terms.

Estimation of probability density function from samples is one of the central problems in statistics and machine learning. Modern neural network-based models can learn high dimensional distributions but have problems with hyperparameter selection and are often prone to instabilities during training and inference. We propose a new efficient tensor train-based model for density estimation (TTDE). Such density parametrization allows exact sampling, calculation of cumulative and marginal density functions, and partition function. It also has very intuitive hyperparameters. We develop an efficient non-adversarial training procedure for TTDE based on the Riemannian optimization. Experimental results demonstrate the competitive performance of the proposed method in density estimation and sampling tasks, while TTDE significantly outperforms competitors in training speed.

Few-shot learning ability is heavily desired for machine intelligence. By meta-learning a model initialization from training tasks with fast adaptation ability to new tasks, model-agnostic meta-learning (MAML) has achieved remarkable success in a number of few-shot learning applications. However, theoretical understandings on the learning ability of MAML remain absent yet, hindering developing new and more advanced meta learning methods in a principled way. In this work, we solve this problem by theoretically justifying the fast adaptation capability of MAML when applied to new tasks. Specifically, we prove that the learnt meta-initialization can benefit the fast adaptation to new tasks with only a few steps of gradient descent. This result explicitly reveals the benefits of the unique designs in MAML. Then we propose a theory-inspired task similarity aware MAML which clusters tasks into multiple groups according to the estimated optimal model parameters and learns group-specific initializations. The proposed method improves upon MAML by speeding up the adaptation and giving stronger few-shot learning ability. Experimental results on the few-shot classification tasks testify its advantages.

Leveraging the framework of Optimal Transport, we introduce a new family of generative autoencoders with a learnable prior, called Symmetric Wasserstein Autoencoders (SWAEs). We propose to symmetrically match the joint distributions of the observed data and the latent representation induced by the encoder and the decoder. The resulting algorithm jointly optimizes the modelling losses in both the data and the latent spaces with the loss in the data space leading to the denoising effect. With the symmetric treatment of the data and the latent representation, the algorithm implicitly preserves the local structure of the data in the latent space. To further improve the quality of the latent representation, we incorporate a reconstruction loss into the objective, which significantly benefits both the generation and reconstruction. We empirically show the superior performance of SWAEs over the state-of-the-art generative autoencoders in terms of classification, reconstruction, and generation.

We propose two sum-product laws for imprecise Markov chains, and use these laws to derive two algorithms to efficiently compute lower and upper expectations for imprecise Markov chains under complete independence and epistemic irrelevance. These algorithms work for inferences that have a corresponding sum-product decomposition, and we argue that many well-known inferences fit their scope. We illustrate our results on a simple epidemiological example.

Deep Gaussian Processes (DGPs) are multi-layer, flexible extensions of Gaussian Processes but their training remains challenging. Most existing methods for inference in DGPs use sparse approximation which require optimization over a large number of inducing inputs and their locations across layers. In this paper, we simplify the training by setting the locations to a fixed subset of data and sampling the inducing inputs from a variational distribution. This reduces the trainable parameters and computation cost without any performance degradation, as demonstrated by our empirical results on regression data sets. Our modifications simplify and stabilize DGP training methods while making them amenable to sampling schemes such as leverage score and determinantal point processes.

Subseasonal climate forecasting is the task of predicting climate variables, such as temperature and precipitation, in a two-week to two-month time horizon. The primary predictors for such prediction problem are spatio-temporal satellite and ground measurements of a variety of climate variables in the atmosphere, ocean, and land, which however have rather limited predictive signal at the subseasonal time horizon. We propose a carefully constructed spatial hierarchical Bayesian regression model that makes use of the inherent spatial structure of the subseasonal climate prediction task. We use our Bayesian model to then derive decision-theoretically optimal point estimates with respect to various performance measures of interest to climate science. As we show, our approach handily improves on various off-the-shelf ML baselines. Since our method is based on a Bayesian framework, we are also able to quantify the uncertainty in our predictions, which is particularly crucial for difficult tasks such as the subseasonal prediction, where we expect any model to have considerable uncertainty at different test locations under different scenarios.

Recent advances in convolutional neural networks (CNNs) usually come with the expense of excessive computational overhead and memory footprint. Network compression aims to alleviate this issue by training compact models with comparable performance. However, existing compression techniques either entail dedicated expert design or compromise with a moderate performance drop. In this paper, we propose a novel structured sparsification method for efficient network compression. The proposed method automatically induces structured sparsity on the convolutional weights, thereby facilitating the implementation of the compressed model with the highly-optimized group convolution. We further address the problem of inter-group communication with a learnable channel shuffle mechanism. The proposed approach can be easily applied to compress many network architectures with a negligible performance drop. Extensive experimental results and analysis demonstrate that our approach gives a competitive performance against the recent network compression counterparts with a sound accuracy-complexity trade-off.

High sample complexity remains a barrier to the application of reinforcement learning (RL), particularly in multi-agent systems. A large body of work has demonstrated that exploration mechanisms based on the principle of optimism under uncertainty can significantly improve the sample efficiency of RL in single agent tasks. This work seeks to understand the role of optimistic exploration in non-cooperative multi-agent settings. We will show that, in zero-sum games, optimistic exploration can cause the learner to waste time sampling parts of the state space that are irrelevant to strategic play, as they can only be reached through cooperation between both players. To address this issue, we introduce a formal notion of strategically efficient exploration in Markov games, and use this to develop two strategically efficient learning algorithms for finite Markov games. We demonstrate that these methods can be significantly more sample efficient than their optimistic counterparts.

We present a novel model for capturing the behavior of an agent exhibiting sunk-cost bias in a stochastic environment. Agents exhibiting sunk-cost bias take into account the effort they have already spent on an endeavor when they evaluate whether to continue or abandon it. We model planning tasks in which an agent with this type of bias tries to reach a designated goal. Our model structures this problem as a type of Markov decision process: loosely speaking, the agent traverses a directed acyclic graph with probabilistic transitions, paying costs for its actions as it tries to reach a target node containing a specified reward. The agent’s sunk cost bias is modeled by a cost that it incurs for abandoning the traversal: if the agent decides to stop traversing the graph, it incurs a cost of $\lambda \cdot C_{sunk}$, where ${\lambda \geq 0}$ is a parameter that captures the extent of the bias and $C_{sunk}$ is the sum of costs already invested. We analyze the behavior of two types of agents: naive agents that are unaware of their bias, and sophisticated agents that are aware of it. Since optimal (bias-free) behavior in this problem can involve abandoning the traversal before reaching the goal, the bias exhibited by these types of agents can result in sub-optimal behavior by shifting their decisions about abandonment. We show that in contrast to optimal agents, it is computationally hard to compute the optimal policy for a sophisticated agent. Our main results quantify the loss exhibited by these two types of agents with respect to an optimal agent. We present both general and topology-specific bounds.