As the class size grows, maintaining a balanced dataset across many classes is challenging because the data are long-tailed in nature; it is even impossible when the sample-of-interest co-exists with each other in one collectable unit, e.g., multiple visual instances in one image. Therefore, long-tailed classification is the key to deep learning at scale. However, existing methods are mainly based on re-weighting/re-sampling heuristics that lack a fundamental theory. In this paper, we establish a causal inference framework, which not only unravels the whys of previous methods, but also derives a new principled solution. Specifically, our theory shows that the SGD momentum is essentially a confounder in long-tailed classification. On one hand, it has a harmful causal effect that misleads the tail prediction biased towards the head. On the other hand, its induced mediation also benefits the representation learning and head prediction. Our framework elegantly disentangles the paradoxical effects of the momentum, by pursuing the direct causal effect caused by an input sample. In particular, we use causal intervention in training, and counterfactual reasoning in inference, to remove the "bad" while keep the "good". We achieve new state-of-the-arts on three long-tailed visual recognition benchmarks: Long-tailed CIFAR-10/-100, ImageNet-LT for image classification and LVIS for instance segmentation.

Low-rank matrix completion recovers incomplete matrix using low-rank assumptions, which is popularly used in many applications. A recent trend is to use nonconvex regularizers that adaptively penalize singular values. They offer good recovery performance and have nice theoretical properties, but are computationally expensive due to repeated access to individual singular values. In this paper, based on the key insight that adaptive shrinkage on singular values improve empirical performance, we propose a new nonconvex low-rank regularizer called "nuclear norm minus Frobenius norm" regularizer, which is scalable, adaptive and sound. We first show it provably holds the adaptive shrinkage property. Further, we discover its factored form which bypasses the computation of singular values and allows fast optimization by general optimization algorithms. Stable recovery and convergence are guaranteed. Extensive experiments on both synthetic and real-world data sets show that the proposed method obtains state-of-the-art performance while being the fastest in comparison to existing methods.

Detecting out-of-distribution (OOD) inputs is critical for safely deploying deep learning models in an open-world setting. However, existing OOD detection solutions can be brittle in the open world, facing various types of adversarial OOD inputs. While methods leveraging auxiliary OOD data have emerged, our analysis reveals a key insight that the majority of auxiliary OOD examples may not meaningfully improve the decision boundary of the OOD detector. In this paper, we provide a theoretically motivated method, Adversarial Training with informative Outlier Mining (ATOM), which improves the robustness of OOD detection. We show that, by mining informative auxiliary OOD data, one can significantly improve OOD detection performance, and somewhat surprisingly, generalize to unseen adversarial attacks. ATOM achieves state-of-the-art performance under a broad family of natural and perturbed OOD evaluation tasks. For example, on the CIFAR-10 in-distribution dataset, ATOM reduces the FPR95 by up to 57.99% under adversarial OOD inputs, surpassing the previous best baseline by a large margin.

High dimensional data is often assumed to be concentrated on or near a low-dimensional manifold. Autoencoders (AE) is a popular technique to learn representations of such data by pushing it through a neural network with a low dimension bottleneck while minimizing a reconstruction error. Using high capacity AE often leads to a large collection of minimizers, many of which represent a low dimensional manifold that fits the data well but generalizes poorly. Two sources of bad generalization are: extrinsic, where the learned manifold possesses extraneous parts that are far from the data; and intrinsic, where the encoder and decoder introduce arbitrary distortion in the low dimensional parameterization. An approach taken to alleviate these issues is to add a regularizer that favors a particular solution; common regularizers promote sparsity, small derivatives, or robustness to noise. In this paper, we advocate an isometry (i.e., local distance preserving) regularizer. Specifically, our regularizer encourages: (i) the decoder to be an isometry; and (ii) the encoder to be the decoder's pseudo-inverse, that is, the encoder extends the inverse of the decoder to the ambient space by orthogonal projection. In a nutshell, (i) and (ii) fix both intrinsic and extrinsic degrees of freedom and provide a non-linear generalization to principal component analysis (PCA). Experimenting with the isometry regularizer on dimensionality reduction tasks produces useful low-dimensional data representations.

Combinations of neural ODEs with recurrent neural networks (RNN), like GRUODE-Bayes or ODE-RNN are well suited to model irregularly-sampled time series. While those models outperform existing discrete-time approaches, no theoretical guarantees for their predictive capabilities are available. Assuming that the irregularly-sampled time series data originates from a continuous stochastic processes, the optimal on-line prediction is the conditional expectation given the currently available information. We introduce the Neural Jump ODE (NJ-ODE) that provides a data-driven approach to learn, continuously in time, the conditional expectation of a stochastic process. Our approach models the conditional expectation between two observations with a neural ODE and jumps whenever a new observation is made. We define a novel training framework, which allows us to prove theoretical convergence guarantees for the first time. In particular, we demonstrate the predictive capabilities of our model by proving that, under some regularity assumptions, the output process converges to the conditional expectation process. We provide experiments showing that the theoretical results also hold empirically. Moreover, we experimentally show that our model outperforms one state of the art model in more complex learning tasks and give comparisons on a real-world dataset.

Learning algorithms are often used in conjunction with expert decision makers in practical scenarios, however this fact is largely ignored when designing these algorithms. In this paper we explore how to learn predictors that can either predict or choose to defer the decision to a downstream expert. Given only samples of the expert's decisions, we give a procedure based on learning a classifier and a rejector and analyze it theoretically. Our approach is based on a novel reduction to cost sensitive learning where we give a consistent surrogate loss for cost sensitive learning that generalizes the cross entropy loss. We show the effectiveness of our approach on a variety of experimental tasks.

We study the statistical and computational properties of a network Lasso method for local graph clustering. The clusters delivered by nLasso can be characterized elegantly via network flows between cluster boundary and seed nodes. While spectral clustering methods are guided by a minimization of the graph Laplacian quadratic form, nLasso minimizes the total variation of cluster indicator signals. As demonstrated theoretically and numerically, nLasso methods can handle very sparse clusters (chain-like) which are difficult for spectral clustering. We also verify that a primal-dual method for nonsmooth optimization allows to approximate nLasso solutions with optimal worst-case convergence rate.

In Bayesian optimization, accounting for the importance of the output relative to the input is a crucial yet challenging exercise, as it can considerably improve the final result but often involves inaccurate and cumbersome entropy estimations. We approach the problem from the perspective of importance-sampling theory, and advocate the use of the likelihood ratio to guide the search algorithm towards regions of the input space where the objective function to be minimized assumes abnormally small values. The likelihood ratio acts as a sampling weight and can be computed at each iteration without severely deteriorating the overall efficiency of the algorithm. In particular, it can be approximated in a way that makes the approach tractable in high dimensions. The "likelihood-weighted" acquisition functions introduced in this work are found to outperform their unweighted counterparts in a number of applications.

Moments when a time series changes its behaviour are called change points. Detection of such points is a well-known problem, which can be found in many applications: quality monitoring of industrial processes, failure detection in complex systems, health monitoring, speech recognition and video analysis. Occurrence of change point implies that the state of the system is altered and its timely detection might help to prevent unwanted consequences. In this paper, we present two online change-point detection approaches based on neural networks. These algorithms demonstrate linear computational complexity and are suitable for change-point detection in large time series. We compare them with the best known algorithms on various synthetic and real world data sets. Experiments show that the proposed methods outperform known approaches.

Recently, the automatic prediction of personality traits has received increasing attention and has emerged as a hot topic within the field of affective computing. In this work, we present a novel deep learning-based approach for automated personality detection from text. We leverage state of the art advances in natural language understanding, namely the BERT language model to extract contextualized word embeddings from textual data for automated author personality detection. Our primary goal is to develop a computationally efficient, high performance personality prediction model which can be easily used by a large number of people without access to huge computation resources. Our extensive experiments with this ideology in mind, led us to develop a novel model which feeds contextualized embeddings along with psycholinguistic features to a Bagged-SVM classifier for personality trait prediction. Our model outperforms the previous state of the art by 1.04% and, at the same time is significantly more computationally efficient to train. We report our results on the famous gold standard Essays dataset for personality detection.