Attention-based encoder decoder network uses a left-to-right beam search algorithm in the inference step. The current beam search expands hypotheses and traverses the expanded hypotheses at the next time step. This traversal is implemented using a for-loop program in general, and it leads to speed down of the recognition process. In this paper, we propose a parallelism technique for beam search, which accelerates the search process by vectorizing multiple hypotheses to eliminate the for-loop program. We also propose a technique to batch multiple speech utterances for off-line recognition use, which reduces the for-loop program with regard to the traverse of multiple utterances. This extension is not trivial during beam search unlike during training due to several pruning and thresholding techniques for efficient decoding. In addition, our method can combine scores of external modules, RNNLM and CTC, in a batch as shallow fusion. We achieved 3.7 x speedup compared with the original beam search algorithm by vectoring hypotheses, and achieved 10.5 x speedup by further changing processing unit to GPU.

Syllabuses for curriculum learning have been developed on an ad-hoc, per task basis and little is known about the relative performance of different syllabuses. We identify a number of syllabuses used in the literature. We compare the identified syllabuses based on their effect on the speed of learning and generalization ability of a LSTM network on three sequential learning tasks. We find that the choice of syllabus has limited effect on the generalization ability of a trained network. In terms of speed of learning our results demonstrate that the best syllabus is task dependent but that a recently proposed automated curriculum learning approach - Predictive Gain, performs very competitively against all identified hand-crafted syllabuses. The best performing hand-crafted syllabus which we term Look Back and Forward combines a syllabus which steps through tasks in the order of their difficulty with a uniform distribution over all tasks. Our experimental results provide an empirical basis for the choice of syllabus on a new problem that could benefit from curriculum learning. Additionally, insights derived from our results shed light on how to successfully design new syllabuses.

Anomaly detection using dimensionality reduction has been an essential technique for monitoring multidimensional data. Although deep learning-based methods have been well studied for their remarkable detection performance, their interpretability is still a problem. In this paper, we propose a novel algorithm for estimating the dimensions contributing to the detected anomalies by using variational autoencoders (VAEs). Our algorithm is based on an approximative probabilistic model that considers the existence of anomalies in the data, and by maximizing the log-likelihood, we estimate which dimensions contribute to determining data as an anomaly. The experiments results with benchmark datasets show that our algorithm extracts the contributing dimensions more accurately than baseline methods.

A learning classifier must outperform a trivial solution, in case of imbalanced data, this condition usually does not hold true. To overcome this problem, we propose a novel data level resampling method - Clustering Based Oversampling for improved learning from class imbalanced datasets. The essential idea behind the proposed method is to use the distance between a minority class sample and its respective cluster centroid to infer the number of new sample points to be generated for that minority class sample. The proposed algorithm has very less dependence on the technique used for finding cluster centroids and does not effect the majority class learning in any way. It also improves learning from imbalanced data by incorporating the distribution structure of minority class samples in generation of new data samples. The newly generated minority class data is handled in a way as to prevent outlier production and overfitting. Implementation analysis on different datasets using deep neural networks as the learning classifier shows the effectiveness of this method as compared to other synthetic data resampling techniques across several evaluation metrics.

Word embedding techniques heavily rely on the abundance of training data for individual words. Given the Zipfian distribution of words in natural language texts, a large number of words do not usually appear frequently or at all in the training data. In this paper we put forward a technique that exploits the knowledge encoded in lexical resources, such as Word-Net, to induce embeddings for unseen words. Our approach adapts graph embedding and cross-lingual vector space transformation techniques in order to merge lexical knowledge encoded in ontologies with that derived from corpus statistics. We show that the approach can provide consistent performance improvements across multiple evaluation benchmarks: in-vitro, on multiple rare word similarity datasets, and invivo, in two downstream text classification tasks.

Abstract--Automated methods for Alzheimer's disease (AD) classification have the potential for great clinical benefits and may provide insight for combating the disease. Machine learning, and more specifically deep neural networks, have been shown to have great efficacy in this domain. These algorithms often use neurological imaging data such as MRI and PET, but a comprehensive and balanced comparison of these modalities has not been performed. In order to accurately determine the relative strength of each imaging variant, this work performs a comparison study in the context of Alzheimer's dementia classification using the Alzheimer's Disease Neuroimaging Initiative (ADNI) Furthermore, this work analyzes the benefits of using both modalities in a fusion setting and discusses how these data types may be leveraged in future AD studies using deep learning. Index Terms--Alzheimer's disease, computer aided diagnosis, convolutional neural network, multimodal, data fusion. LZHEIMER'S disease (AD) is a neurodegenerative disorder characterized by cognitive decline and dementia. The number of individuals living with AD in the United States is expected to reach 10 million by the year 2025 [1]. As a result, automated methods for computer aided diagnosis could greatly improve the ability to screen at-risk individuals. Such methods typically take as input patient data including demographics, medical history, genetic sequencing, and neurological images among others. The resulting output is health status indicated by a diagnosis label, which may also include a probabilistic uncertainty on the prediction.

We introduce the variational filtering EM algorithm, a simple, general-purpose method for performing variational inference in dynamical latent variable models using information from only past and present variables, i.e. filtering. The algorithm is derived from the variational objective in the filtering setting and consists of an optimization procedure at each time step. By performing each inference optimization procedure with an iterative amortized inference model, we obtain a computationally efficient implementation of the algorithm, which we call amortized variational filtering. We present experiments demonstrating that this general-purpose method improves performance across several deep dynamical latent variable models.

The soaring demand for intelligent mobile applications calls for deploying powerful deep neural networks (DNNs) on mobile devices. However, the outstanding performance of DNNs notoriously relies on increasingly complex models, which in turn is associated with an increase in computational expense far surpassing mobile devices' capacity. What is worse, app service providers need to collect and utilize a large volume of users' data, which contain sensitive information, to build the sophisticated DNN models. Directly deploying these models on public mobile devices presents prohibitive privacy risk. To benefit from the on-device deep learning without the capacity and privacy concerns, we design a private model compression framework RONA. Following the knowledge distillation paradigm, we jointly use hint learning, distillation learning, and self learning to train a compact and fast neural network. The knowledge distilled from the cumbersome model is adaptively bounded and carefully perturbed to enforce differential privacy. We further propose an elegant query sample selection method to reduce the number of queries and control the privacy loss. A series of empirical evaluations as well as the implementation on an Android mobile device show that RONA can not only compress cumbersome models efficiently but also provide a strong privacy guarantee. For example, on SVHN, when a meaningful $(9.83,10^{-6})$-differential privacy is guaranteed, the compact model trained by RONA can obtain 20$\times$ compression ratio and 19$\times$ speed-up with merely 0.97% accuracy loss.

The use of deep neural networks in edge computing devices hinges on the balance between accuracy and complexity of computations. Ternary Connect (TC) \cite{lin2015neural} addresses this issue by restricting the parameters to three levels $-1, 0$, and $+1$, thus eliminating multiplications in the forward pass of the network during prediction. We propose Generalized Ternary Connect (GTC), which allows an arbitrary number of levels while at the same time eliminating multiplications by restricting the parameters to integer powers of two. The primary contribution is that GTC learns the number of levels and their values for each layer, jointly with the weights of the network in an end-to-end fashion. Experiments on MNIST and CIFAR-10 show that GTC naturally converges to an `almost binary' network for deep classification networks (e.g. VGG-16) and deep variational auto-encoders, with negligible loss of classification accuracy and comparable visual quality of generated samples respectively. We demonstrate superior compression and similar accuracy of GTC in comparison to several state-of-the-art methods for neural network compression. We conclude with simulations showing the potential benefits of GTC in hardware.

Statistical Machine Learning (SML) refers to a body of algorithms and methods by which computers are allowed to discover important features of input data sets which are often very large in size. The very task of feature discovery from data is essentially the meaning of the keyword `learning' in SML. Theoretical justifications for the effectiveness of the SML algorithms are underpinned by sound principles from different disciplines, such as Computer Science and Statistics. The theoretical underpinnings particularly justified by statistical inference methods are together termed as statistical learning theory. This paper provides a review of SML from a Bayesian decision theoretic point of view -- where we argue that many SML techniques are closely connected to making inference by using the so called Bayesian paradigm. We discuss many important SML techniques such as supervised and unsupervised learning, deep learning, online learning and Gaussian processes especially in the context of very large data sets where these are often employed. We present a dictionary which maps the key concepts of SML from Computer Science and Statistics. We illustrate the SML techniques with three moderately large data sets where we also discuss many practical implementation issues. Thus the review is especially targeted at statisticians and computer scientists who are aspiring to understand and apply SML for moderately large to big data sets.