This paper introduces ASCAI, a novel adaptive sampling methodology that can learn how to effectively compress Deep Neural Networks (DNNs) for accelerated inference on resource-constrained platforms. Modern DNN compression techniques comprise various hyperparameters that require per-layer customization to ensure high accuracy. Choosing such hyperparameters is cumbersome as the pertinent search space grows exponentially with the number of model layers. To effectively traverse this large space, we devise an intelligent sampling mechanism that adapts the sampling strategy using customized operations inspired by genetic algorithms. As a special case, we consider the space of model compression as a vector space. The adaptively selected samples enable ASCAI to automatically learn how to tune per-layer compression hyperparameters to optimize the accuracy/model-size trade-off. Our extensive evaluations show that ASCAI outperforms rule-based and reinforcement learning methods in terms of compression rate and/or accuracy

Rademacher complexity plays a fundamental role in learning theory, where it tightly bounds the supremum of the empirical process ( Koltchinskii and Panchenko, 2000; Bartlett and Mendelson, 2003) and is used to prove generalization guarantees for empiric al risk minimization and other learning rules.

Jerry Quinn IBM T.J. Watson Research Center Y orktown Heights, NY 10598 V alentina Salapura IBM T.J. Watson Research Center Y orktown Heights, NY 10598 Abstract This paper presents a methodology for selecting the mini-batch size that minimizes Stochastic Gradient Descent (SGD) learning time for single and multiple learner problems. By de-coupling algorithmic analysis issues from hardware and software implementation details, we reveal a robust empirical inverse law between mini-batch size and the average number of SGD updates required to converge to a specified error threshold. Combining this empirical inverse law with measured system performance, we create an accurate, closed-form model of average training time and show how this model can be used to identify quantifiable implications for both algorithmic and hardware aspects of machine learning. We demonstrate the inverse law empirically, on both image recognition (MNIST, CIFAR10 and CIFAR100) and machine translation (Europarl) tasks, and provide a theoretic justification via proving a novel bound on mini-batch SGD training. Introduction In this paper, we present an empirical law, with theoretical justification, linking the number of learning iterations to the mini-batch size. From this result, we derive a principled methodology for selecting mini-batch size w.r.t. This methodology saves training time and provides both intuition and a principled approach for optimizing machine learning algorithms and machine learning hardware system design. Further, we use our methodology to show that focusing on weak scaling can lead to suboptimal training times because, by neglecting the dependence of convergence time on the size of the mini-batch used, weak scaling does not always minimize the training time.

Time series data are prevalent in electronic health records, mostly in the form of physiological parameters such as vital signs and lab tests. The patterns of these values may be significant indicators of patients' clinical states and there might be patterns that are unknown to clinicians but are highly predictive of some outcomes. Many of these values are also missing which makes it difficult to apply existing methods like decision trees. We propose a recurrent neural network model that reduces overfitting to noisy observations by limiting interactions between features. We analyze its performance on mortality, ICD-9 and AKI prediction from observational values on the Medical Information Mart for Intensive Care III (MIMIC-III) dataset. Our models result in an improvement of 1.1% [p<0.01] in AU-ROC for mortality prediction under the MetaVision subset and 1.0% and 2.2% [p<0.01] respectively for mortality and AKI under the full MIMIC-III dataset compared to existing state-of-the-art interpolation, embedding and decay-based recurrent models.

We present ProxiModel, a novel event mining framework for extracting high-quality structured event knowledge from large, redundant, and noisy news data sources. The proposed model differentiates itself from other approaches by modeling both the event correlation within each individual document as well as across the corpus. To facilitate this, we introduce the concept of a proximity-network, a novel space-efficient data structure to facilitate scalable event mining. This proximity network captures the corpus-level co-occurence statistics for candidate event descriptors, event attributes, as well as their connections. We probabilistically model the proximity network as a generative process with sparsity-inducing regularization. This allows us to efficiently and effectively extract high-quality and interpretable news events. Experiments on three different news corpora demonstrate that the proposed method is effective and robust at generating high-quality event descriptors and attributes. We briefly detail many interesting applications from our proposed framework such as news summarization, event tracking and multi-dimensional analysis on news. Finally, we explore a case study on visualizing the events for a Japan Tsunami news corpus and demonstrate ProxiModel's ability to automatically summarize emerging news events.

Convolutional neural networks (CNNs) with dilated filters such as the Wavenet or the Temporal Convolutional Network (TCN) have shown good results in a variety of sequence modelling tasks. However, efficiently modelling long-term dependencies in these sequences is still challenging. Although the receptive field of these models grows exponentially with the number of layers, computing the convolutions over very long sequences of features in each layer is time and memory-intensive, prohibiting the use of longer receptive fields in practice. To increase efficiency, we make use of the "slow feature" hypothesis stating that many features of interest are slowly varying over time. For this, we use a U-Net architecture that computes features at multiple time-scales and adapt it to our auto-regressive scenario by making convolutions causal. We apply our model ("Seq-U-Net") to a variety of tasks including language and audio generation. In comparison to TCN and Wavenet, our network consistently saves memory and computation time, with speed-ups for training and inference of over 4x in the audio generation experiment in particular, while achieving a comparable performance in all tasks.

In this paper we address the problem of solving ill-posed inverse problems in imaging where the prior is a neural generative model. Specifically we consider the decoupled case where the prior is trained once and can be reused for many different log-concave degradation models without retraining. Whereas previous MAP-based approaches to this problem lead to highly non-convex optimization algorithms, our approach computes the joint (space-latent) MAP that naturally leads to alternate optimization algorithms and to the use of a stochastic encoder to accelerate computations. The resulting technique is called JPMAP because it performs Joint Posterior Maximization using an Autoencoding Prior. We show theoretical and experimental evidence that the proposed objective function is quite close to bi-convex. Indeed it satisfies a weak bi-convexity property which is sufficient to guarantee that our optimization scheme converges to a stationary point. Experimental results also show the higher quality of the solutions obtained by our JPMAP approach with respect to other non-convex MAP approaches which more often get stuck in spurious local optima.

Adverse drug events (ADEs) are "injuries resulting from medical intervention related to a drug" (Nebeker, Barach, and Samore 2004), and are distinct from medication errors (inappropriate prescription, dispensing, usage etc.) as they are caused by drugs at normal dosages. According to the National Center for Health Statistics (NCHS 2014), 48.9% of Americans took at least one prescription drug in the last 30 days, 23.1% took at least three, and 11.9% took at least

Multi-output Gaussian processes (MOGPs) leverage the flexibility and interpretability of GPs while capturing structure across outputs, which is desirable, for example, in spatio-temporal modelling. The key problem with MOGPs is the cubic computational scaling in the number of both inputs (e.g., time points or locations), n, and outputs, p. Current methods reduce this to O(n^3 m^3), where m < p is the desired degrees of freedom. This computational cost, however, is still prohibitive in many applications. To address this limitation, we present the Orthogonal Linear Mixing Model (OLMM), an MOGP in which exact inference scales linearly in m: O(n^3 m). This advance opens up a wide range of real-world tasks and can be combined with existing GP approximations in a plug-and-play way as demonstrated in the paper. Additionally, the paper organises the existing disparate literature on MOGP models into a simple taxonomy called the Mixing Model Hierarchy (MMH).

We propose a regression algorithm that utilizes a learned dictionary optimized for sparse inference on D-Wave quantum annealer. In this regression algorithm, we concatenate the independent and dependent variables as an combined vector, and encode the high-order correlations between them into a dictionary optimized for sparse reconstruction. On a test dataset, the dependent variable is initialized to its average value and then a sparse reconstruction of the combined vector is obtained in which the dependent variable is typically shifted closer to its true value, as in a standard inpainting or denoising task. Here, a quantum annealer, which can presumably exploit a fully entangled initial state to better explore the complex energy landscape, is used to solve the highly non-convex sparse coding optimization problem. The regression algorithm is demonstrated for a lattice quantum chromodynamics simulation data using a D-Wave 2000Q quantum annealer and good prediction performance is achieved. The regression test is performed using six different values for the number of fully connected logical qubits, between 20 and 64, the latter being the maximum that can be embedded on the D-Wave 2000Q. The scaling results indicate that a larger number of qubits gives better prediction accuracy, the best performance being comparable to the best classical regression algorithms reported so far.