Learning long-term dependencies is a key long-standing challenge of recurrent neural networks (RNNs). Hierarchical recurrent neural networks (HRNNs) have been considered a promising approach as long-term dependencies are resolved through shortcuts up and down the hierarchy. Yet, the memory requirements of Truncated Backpropagation Through Time (TBPTT) still prevent training them on very long sequences. In this paper, we empirically show that in (deep) HRNNs, propagating gradients back from higher to lower levels can be replaced by locally computable losses, without harming the learning capability of the network, over a wide range of tasks. This decoupling by local losses reduces the memory requirements of training by a factor exponential in the depth of the hierarchy in comparison to standard TBPTT.

Traditional sequential multi-object attention models rely on a recurrent mechanism to infer object relations. We propose a relational extension (R-SQAIR) of one such attention model (SQAIR) by endowing it with a module with strong relational inductive bias that computes in parallel pairwise interactions between inferred objects. Two recently proposed relational modules are studied on tasks of unsupervised learning from videos. We demonstrate gains over sequential relational mechanisms, also in terms of combinatorial generalization.

Here, we use the nomenclature derived from the meta-learning literature which is the most prevalent at time of writing. Let S {( x l,y l)} L l 1 be a support-set containing input-output pairs, with L equal to one (1-shot) or five (5-shot), and Q { (x m,y m)} M m 1be a query-set (sometimes referred to in the literature as a target-set), with M typically one order of magnitude greater than L. For ease of notation, the support and query sets are grouped in a task T {S, Q}, with the dataset D {T n} N n 1 defined as a collection of such tasks. Models are trained on random tasks sampled from D . Then, given a new task T {S, Q } sampled from a test set, the objective is to condition the model on the samples of the support S to estimate the membership of the samples in the query set Q . In the most common scenario, the inputs x D belong to the same distribution p(x) and are distributed across training, validation, and test sets such that their class membership is non-overlapping. Note that y can be a continuous value (regression) or a discrete one (classification), even though most of the previous work has focused on classification. We also consider the cross-domain scenario, where the inputs are sampled from different distributions at training and test time; this is more representative of real-world scenarios.

We investigate the theoretical limits of pipeline parallel learning of deep learning architectures, a distributed setup in which the computation is distributed per layer instead of per example. For smooth convex and non-convex objective functions, we provide matching lower and upper complexity bounds and show that a naive pipeline parallelization of Nesterov's accelerated gradient descent is optimal. For non-smooth convex functions, we provide a novel algorithm coined Pipeline Parallel Random Smoothing (PPRS) that is within a $d^{1/4}$ multiplicative factor of the optimal convergence rate, where $d$ is the underlying dimension. While the convergence rate still obeys a slow $\varepsilon^{-2}$ convergence rate, the depth-dependent part is accelerated, resulting in a near-linear speed-up and convergence time that only slightly depends on the depth of the deep learning architecture. Finally, we perform an empirical analysis of the non-smooth non-convex case and show that, for difficult and highly non-smooth problems, PPRS outperforms more traditional optimization algorithms such as gradient descent and Nesterov's accelerated gradient descent for problems where the sample size is limited, such as few-shot or adversarial learning.

Inspired by trial-to-trial variability in the brain that can result from multiple noise sources, we introduce variability through noise at different levels in a knowledge distillation framework. We introduce "Fickle Teacher" which provides variable supervision signals to the student for the same input. We observe that the response variability from the teacher results in a significant generalization improvement in the student. We further propose "Soft-Randomization" as a novel technique for improving robustness to input variability in the student. This minimizes the dissimilarity between the student's distribution on noisy data with teacher's distribution on clean data. We show that soft-randomization, even with low noise intensity, improves the robustness significantly with minimal drop in generalization. Lastly, we propose a new technique, "Messy-collaboration", which introduces target variability, whereby student and/or teacher are trained with randomly corrupted labels. We find that supervision from a corrupted teacher improves the adversarial robustness of student significantly while preserving its generalization and natural robustness. Our extensive empirical results verify the effectiveness of adding constructive noise in the knowledge distillation framework for improving the generalization and robustness of the model.

Dynamical system models (including RNNs) often lack the ability to adapt the sequence generation or prediction to a given context, limiting their real-world application. In this paper we show that hierarchical multi-task dynamical systems (MTDSs) provide direct user control over sequence generation, via use of a latent code $\mathbf{z}$ that specifies the customization to the individual data sequence. This enables style transfer, interpolation and morphing within generated sequences. We show the MTDS can improve predictions via latent code interpolation, and avoid the long-term performance degradation of standard RNN approaches.

The evolution of social media users' behavior over time complicates user-level comparison tasks such as verification, classification, clustering, and ranking. As a result, na ıve approaches may fail to generalize to new users or even to future observations of previously known users. In this paper, we propose a novel procedure to learn a mapping from short episodes of user activity on social media to a vector space in which the distance between points captures the similarity of the corresponding users' invariant features. We fit the model by optimizing a surrogate metric learning objective over a large corpus of unlabeled social media content. Once learned, the mapping may be applied to users not seen at training time and enables efficient comparisons of users in the resulting vector space. We present a comprehensive evaluation to validate the benefits of the proposed approach using data from Reddit, Twitter, and Wikipedia.

Department of Computer Science and Engineering Texas A&M University College Station, TX 77840, USA Abstract PyODDS is an end-to-end Py thon system for O utlier D etection with Database Support. It provides various outlier detection algorithms which meet the demands for users in different fields, with or without data science or machine learning background. PyODDS gives the ability to execute machine learning algorithms in-database without moving data out of the database server or over the network. It also provides access to a wide range of outlier detection algorithms, including statistical analysis and more recent deep learning based approaches. Keywords: anomaly detection, end-to-end system, outlier detection, deep learning, machine learning, data mining, full stack system, data visualization 1. Introduction Outliers refer to the objects with patterns or behaviors that are significantly rare and different with the rest of majorities.

ABSTRACT There is an implicit assumption that traditional hybrid approaches for automatic speech recognition (ASR) cannot directly model graphemes and need to rely on phonetic lexicons to get competitive performance, especially on English which has poor grapheme-phoneme correspondence. In this work, we show for the first time that, on English, hybrid ASR systems can in fact model graphemes effectively by leveraging tied context-dependent graphemes, i.e., chenones. Our chenone-based systems significantly outperform equivalent senone baselines by 4.5% to 11.1% relative on three different English datasets. Our results on Librispeech are state-of- the-art compared to other hybrid approaches and competitive with previously published end-to-end numbers. Further analysis shows that chenones can better utilize powerful acoustic models and large training data, and require context-and position-dependent modeling to work well. Chenone-based systems also outperform senone baselines on proper noun and rare word recognition, an area where the latter is traditionally thought to have an advantage. Our work provides an alternative for end-to-end ASR and establishes that hybrid systems can be improved by dropping the reliance on phonetic knowledge. Index T erms-- graphemic lexicon, hybrid speech recognition, chenones, acoustic modeling, librispeech 1. INTRODUCTION In the past decade, neural network acoustic models have become a staple in automatic speech recognition (ASR).

PoPPy is a Point Process toolbox based on PyTorch, which achieves flexible designing and efficient learning of point process models. It can be used for interpretable sequential data modeling and analysis, e.g., Granger causality analysis of multi-variate point processes, point process-based simulation and prediction of event sequences. In practice, the key points of point process-based sequential data modeling include: 1) How to design intensity functions to describe the mechanism behind observed data? 2) How to learn the proposed intensity functions from observed data? The goal of PoPPy is providing a user-friendly solution to the key points above and achieving large-scale point process-based sequential data analysis, simulation and prediction.