The field of language modelling has been largely dominated by autoregressive models, for which sampling is inherently difficult to parallelize. This paper introduces two new classes of generative models for categorical data such as language or image segmentation: Argmax Flows and Multinomial Diffusion. Argmax Flows are defined by a composition of a continuous distribution (such as a normalizing flow), and an argmax function. To optimize this model, we learn a probabilistic inverse for the argmax that lifts the categorical data to a continuous space. Multinomial Diffusion gradually adds categorical noise in a diffusion process, for which the generative denoising process is learned. We demonstrate that our models perform competitively on language modelling and modelling of image segmentation maps.

A salient feature of a over-parameterized model, e.g. a deep neural network, is its ability to memorize examples Zhang et al. (2016); Neyshabur et al. (2017), and the memorization has proven to benefit the generalization Arpit et al. (2017); Feldman (2020); Feldman & Zhang (2020). Nonetheless, the potential existence of label noise, combined with the memorization effect, might lead to detrimental consequence Song et al. (2020); Yao et al. (2020a); Cheng et al. (2020b); Chen et al. (2019); Han et al. (2020); Song et al. (2020). In light of the reported empirical evidence of harms caused by over-memorizing noisy labels, we set out to understand this effect analytically. Built on a recent analytical framework Feldman (2020), we demonstrate the varying effects of memorizing noisy labels associated with instances that sit at the different spectra of the sample distribution (Section 3). Soon since the above negative effect was empirically shown, learning with noisy labels has been recognized as a challenging and important task.

We consider nonnegative time series forecasting framework. Based on recent advances in Nonnegative Matrix Factorization (NMF) and Archetypal Analysis, we introduce two procedures referred to as Sliding Mask Method (SMM) and Latent Clustered Forecast (LCF). SMM is a simple and powerful method based on time window prediction using Completion of Nonnegative Matrices. This new procedure combines low nonnegative rank decomposition and matrix completion where the hidden values are to be forecasted. LCF is two stage: it leverages archetypal analysis for dimension reduction and clustering of time series, then it uses any black-box supervised forecast solver on the clustered latent representation. Theoretical guarantees on uniqueness and robustness of the solution of NMF Completion-type problems are also provided for the first time. Finally, numerical experiments on real-world and synthetic data-set confirms forecasting accuracy for both the methodologies.

Granger causal modeling is an emerging topic that can uncover Granger causal relationship behind multivariate time series data. In many real-world systems, it is common to encounter a large amount of multivariate time series data collected from different individuals with sharing commonalities. However, there are ongoing concerns regarding Granger causality's applicability in such large scale complex scenarios, presenting both challenges and opportunities for Granger causal structure reconstruction. Existing methods usually train a distinct model for each individual, suffering from inefficiency and over-fitting issues. To bridge this gap, we propose an Inductive GRanger cAusal modeling (InGRA) framework for inductive Granger causality learning and common causal structure detection on multivariate time series, which exploits the shared commonalities underlying the different individuals. In particular, we train one global model for individuals with different Granger causal structures through a novel attention mechanism, called prototypical Granger causal attention. The model can detect common causal structures for different individuals and infer Granger causal structures for newly arrived individuals. Extensive experiments, as well as an online A/B test on an E-commercial advertising platform, demonstrate the superior performances of InGRA.

Model-based reinforcement learning (MBRL) approaches rely on discrete-time state transition models whereas physical systems and the vast majority of control tasks operate in continuous-time. To avoid time-discretization approximation of the underlying process, we propose a continuous-time MBRL framework based on a novel actor-critic method. Our approach also infers the unknown state evolution differentials with Bayesian neural ordinary differential equations (ODE) to account for epistemic uncertainty. We implement and test our method on a new ODE-RL suite that explicitly solves continuous-time control systems. Our experiments illustrate that the model is robust against irregular and noisy data, is sample-efficient, and can solve control problems which pose challenges to discrete-time MBRL methods.

The recent success of supervised learning methods on ever larger offline datasets has spurred interest in the reinforcement learning (RL) field to investigate whether the same paradigms can be translated to RL algorithms. This research area, known as offline RL, has largely focused on offline policy optimization, aiming to find a return-maximizing policy exclusively from offline data. In this paper, we consider a slightly different approach to incorporating offline data into sequential decision-making. We aim to answer the question, what unsupervised objectives applied to offline datasets are able to learn state representations which elevate performance on downstream tasks, whether those downstream tasks be online RL, imitation learning from expert demonstrations, or even offline policy optimization based on the same offline dataset? Through a variety of experiments utilizing standard offline RL datasets, we find that the use of pretraining with unsupervised learning objectives can dramatically improve the performance of policy learning algorithms that otherwise yield mediocre performance on their own. Extensive ablations further provide insights into what components of these unsupervised objectives -- e.g., reward prediction, continuous or discrete representations, pretraining or finetuning -- are most important and in which settings.

Inspired by the inductive transfer learning on computer vision, many efforts have been made to train contextualized language models that boost the performance of natural language processing tasks. These models are mostly trained on large general-domain corpora such as news, books, or Wikipedia.Although these pre-trained generic language models well perceive the semantic and syntactic essence of a language structure, exploiting them in a real-world domain-specific scenario still needs some practical considerations to be taken into account such as token distribution shifts, inference time, memory, and their simultaneous proficiency in multiple tasks. In this paper, we focus on the legal domain and present how different language model strained on general-domain corpora can be best customized for multiple legal document reviewing tasks. We compare their efficiencies with respect to task performances and present practical considerations.

Reinforcement learning is about learning agent models that make the best sequential decisions in unknown environments. In an unknown environment, the agent needs to explore the environment while exploiting the collected information, which usually forms a sophisticated problem to solve. Derivative-free optimization, meanwhile, is capable of solving sophisticated problems. It commonly uses a sampling-and-updating framework to iteratively improve the solution, where exploration and exploitation are also needed to be well balanced. Therefore, derivative-free optimization deals with a similar core issue as reinforcement learning, and has been introduced in reinforcement learning approaches, under the names of learning classifier systems and neuroevolution/evolutionary reinforcement learning. Although such methods have been developed for decades, recently, derivative-free reinforcement learning exhibits attracting increasing attention. However, recent survey on this topic is still lacking. In this article, we summarize methods of derivative-free reinforcement learning to date, and organize the methods in aspects including parameter updating, model selection, exploration, and parallel/distributed methods. Moreover, we discuss some current limitations and possible future directions, hoping that this article could bring more attentions to this topic and serve as a catalyst for developing novel and efficient approaches.

We present an efficient foveal framework to perform object detection. A scale normalized image pyramid (SNIP) is generated that, like human vision, only attends to objects within a fixed size range at different scales. Such a restriction of objects' size during training affords better learning of object-sensitive filters, and therefore, results in better accuracy. However, the use of an image pyramid increases the computational cost. Hence, we propose an efficient spatial sub-sampling scheme which only operates on fixed-size sub-regions likely to contain objects (as object locations are known during training). The resulting approach, referred to as Scale Normalized Image Pyramid with Efficient Resampling or SNIPER, yields up to 3 times speed-up during training. Unfortunately, as object locations are unknown during inference, the entire image pyramid still needs processing. To this end, we adopt a coarse-to-fine approach, and predict the locations and extent of object-like regions which will be processed in successive scales of the image pyramid. Intuitively, it's akin to our active human-vision that first skims over the field-of-view to spot interesting regions for further processing and only recognizes objects at the right resolution. The resulting algorithm is referred to as AutoFocus and results in a 2.5-5 times speed-up during inference when used with SNIP.

A core challenge in Machine Learning is to learn to disentangle natural factors of variation in data (e.g. object shape vs. pose). A popular approach to disentanglement consists in learning to map each of these factors to distinct subspaces of a model's latent representation. However, this approach has shown limited empirical success to date. Here, we show that, for a broad family of transformations acting on images--encompassing simple affine transformations such as rotations and translations--this approach to disentanglement introduces topological defects (i.e. discontinuities in the encoder). Motivated by classical results from group representation theory, we study an alternative, more flexible approach to disentanglement which relies on distributed latent operators, potentially acting on the entire latent space. We theoretically and empirically demonstrate the effectiveness of this approach to disentangle affine transformations. Our work lays a theoretical foundation for the recent success of a new generation of models using distributed operators for disentanglement.