Pseudo-relevance feedback (PRF) is commonly used to boost the performance of traditional information retrieval (IR) models by using top-ranked documents to identify and weight new query terms, thereby reducing the effect of query-document vocabulary mismatches. While neural retrieval models have recently demonstrated strong results for ad-hoc retrieval, combining them with PRF is not straightforward due to incompatibilities between existing PRF approaches and neural architectures. To bridge this gap, we propose an end-to-end neural PRF framework that can be used with existing neural IR models by embedding different neural models as building blocks. Extensive experiments on two standard test collections confirm the effectiveness of the proposed NPRF framework in improving the performance of two state-of-the-art neural IR models.

Transfer Learning (TL) in Deep Neural Networks is gaining importance because in most of the cases, the labeling of data is costly and time-consuming. Additionally, TL provides effective weight initialization. This paper introduces the idea of Adaptive Transfer Learning in Deep Neural Networks (ATL-DNN) for wind power prediction. Adaptive TL of Deep Neural Networks is proposed, which makes the proposed system an adaptive one as regards training on a different wind farm is concerned. The proposed ATL-DNN technique is tested for short-term wind power predictions, where continuously arriving information has to be exploited. Adaptive TL not only helps in providing good weight initialization, but is also helpful to utilize the online data that is continuously being generated by wind farms. Additionally, the proposed ATL-DNN technique is shown to transfer knowledge between different task domains (wind power to wind speed prediction) and from one region to another region. The simulation results show that proposed ATL-DNN technique achieves average values of 0.0637,0.0986, Keywords ---- Wind power prediction; Adaptive transfer learning; Deep learning; Ensemble learning 1. Introduction Many countries across the world use wind power as a renewable energy resource. Accurate prediction of wind power plays a significant role in generating smooth power from a turbine. There are numerous factors which affect the predicted power of a wind power prediction system, like fluctuation in speed of the wind with respect to time, geographical location, and climatic conditions.

ABSTRACT The state-of-the-art in text-to-speech synthesis has recently improved considerably due to novel neural waveform generation methods, such as WaveNet. However, these methods suffer from their slow sequential inference process, while their parallel versions are difficult to train and even more expensive computationally. Meanwhile, generative adversarial networks (GANs) have achieved impressive results in image generation and are making their way into audio applications; parallel inference is among their lucrative properties. Listening test results show that while direct waveform generation with GAN is still far behind WaveNet, a GAN-based glottal excitation model can achieve quality and voice similarity on par with a WaveNet vocoder. Index Terms-- Neural vocoding, text-to-speech, GAN, glottal excitation model 1. INTRODUCTION Recent advances in deep learning have led to text-to-speech (TTS) systems achieving near-human naturalness [1]. This is partially due to neural sequence-to-sequence mapping methods that can learn to align and map between input text and output acoustic feature sequences [2].

We advocate Laplacian K-modes for joint clustering and density mode finding, and propose a concave-convex relaxation of the problem, which yields a parallel algorithm that scales up to large datasets and high dimensions. We optimize a tight bound (auxiliary function) of our relaxation, which, at each iteration, amounts to computing an independent update for each cluster-assignment variable, with guaranteed convergence. Therefore, our bound optimizer can be trivially distributed for large-scale data sets. Furthermore, we show that the density modes can be obtained as byproducts of the assignment variables via simple maximum-value operations whose additional computational cost is linear in the number of data points. Our formulation does not need storing a full affinity matrix and computing its eigenvalue decomposition, neither does it perform expensive projection steps and Lagrangian-dual inner iterates for the simplex constraints of each point. Furthermore, unlike mean-shift, our density-mode estimation does not require inner-loop gradient-ascent iterates. It has a complexity independent of feature-space dimension, yields modes that are valid data points in the input set and is applicable to discrete domains as well as arbitrary kernels. We report comprehensive experiments over various data sets, which show that our algorithm yields very competitive performances in term of optimization quality (i.e., the value of the discrete-variable objective at convergence) and clustering accuracy.

Conditional Density Estimation (CDE) models deal with estimating conditional distributions. The conditions imposed on the distribution are the inputs of the model. CDE is a challenging task as there is a fundamental trade-off between model complexity, representational capacity and overfitting. In this work, we propose to extend the model's input with latent variables and use Gaussian processes (GP) to map this augmented input onto samples from the conditional distribution. Our Bayesian approach allows for the modeling of small datasets, but we also provide the machinery for it to be applied to big data using stochastic variational inference. Our approach can be used to model densities even in sparse data regions, and allows for sharing learned structure between conditions. We illustrate the effectiveness and wide-reaching applicability of our model on a variety of real-world problems, such as spatio-temporal density estimation of taxi drop-offs, non-Gaussian noise modeling, and few-shot learning on omniglot images.

Scientific innovations relating physical processes require laborious experimentation and expensive simulation as the relationships between design variables and output characteristics are unknown [4, 28]. To design a new product with certain target characteristics, a search is typically performed in the design space - a large number of the design combinations (input variables) are tried in simulators before reaching to the target characteristics (see Figure 1a). Although modern search methods such as Bayesian Optimization [30] are efficient, there is an inherent problem with the search: Each time the target characteristics are changed or refined, the search process needs to be restarted making the design task extremely time-consuming. The current search paradigm does not harness existing experimentation data and simulator queries systematically as there is no clear provision to reuse them. Thanks to the availability of growing data, accurate simulators and modern machine learning algorithms, this search process can be avoided with a potential to accelerate scientific innovations by multiple orders of magnitude. An effective way to address this problem is to leverage the existing data and query the simulators in an offline mode to sample the data space sufficiently and then harness this data to build a machine learning model. Given sufficient data, modern machine learning (ML) models (e.g. a deep neural network) can approximate the underlying physical relationships and the simulators arbitrarily closely. The ML models can then be used to convert the search process in a less expensive optimization.

Recommender systems help users find relevant items of interest, for example on e-commerce or media streaming sites. Most academic research is concerned with approaches that personalize the recommendations according to long-term user profiles. In many real-world applications, however, such long-term profiles often do not exist and recommendations therefore have to be made solely based on the observed behavior of a user during an ongoing session. Given the high practical relevance of the problem, an increased interest in this problem can be observed in recent years, leading to a number of proposals for session-based recommendation algorithms that typically aim to predict the user's immediate next actions. In this work, we present the results of an in-depth performance comparison of a number of such algorithms, using a variety of datasets and evaluation measures. Our comparison includes the most recent approaches based on recurrent neural networks like GRU4REC, factorized Markov model approaches such as FISM or FOSSIL, as well as simpler methods based, e.g., on nearest neighbor schemes. Our experiments reveal that algorithms of this latter class, despite their sometimes almost trivial nature, often perform equally well or significantly better than today's more complex approaches based on deep neural networks. Our results therefore suggest that there is substantial room for improvement regarding the development of more sophisticated session-based recommendation algorithms.

Recent studies have shown promising results in using Deep Learning to detect malignancy in whole slide imaging. However, they were limited to just predicting positive or negative finding for a specific neoplasm. We attempted to use Deep Learning with a convolutional neural network algorithm to build a lymphoma diagnostic model for four diagnostic categories: benign lymph node, diffuse large B cell lymphoma, Burkitt lymphoma, and small lymphocytic lymphoma. Our software was written in Python language. We obtained digital whole slide images of Hematoxylin and Eosin stained slides of 128 cases including 32 cases for each diagnostic category. Four sets of 5 representative images, 40x40 pixels in dimension, were taken for each case. A total of 2,560 images were obtained from which 1,856 were used for training, 464 for validation and 240 for testing. For each test set of 5 images, the predicted diagnosis was combined from prediction of 5 images. The test results showed excellent diagnostic accuracy at 95% for image-by-image prediction and at 10% for set-by-set prediction. This preliminary study provided a proof of concept for incorporating automated lymphoma diagnostic screen into future pathology workflow to augment the pathologists' productivity.

With an aging and growing population, the number of women requiring either screening or symptomatic mammograms is increasing. To reduce the number of mammograms that need to be read by a radiologist while keeping the diagnostic accuracy the same or better than current clinical practice, we develop Man and Machine Mammography Oracle (MAMMO) - a clinical decision support system capable of triaging mammograms into those that can be confidently classified by a machine and those that cannot be, thus requiring the reading of a radiologist. The first component of MAMMO is a novel multi-view convolutional neural network (CNN) with multi-task learning (MTL). MTL enables the CNN to learn the radiological assessments known to be associated with cancer, such as breast density, conspicuity, suspicion, etc., in addition to learning the primary task of cancer diagnosis. We show that MTL has two advantages: 1) learning refined feature representations associated with cancer improves the classification performance of the diagnosis task and 2) issuing radiological assessments provides an additional layer of model interpretability that a radiologist can use to debug and scrutinize the diagnoses provided by the CNN. The second component of MAMMO is a triage network, which takes as input the radiological assessment and diagnostic predictions of the first network's MTL outputs and determines which mammograms can be correctly and confidently diagnosed by the CNN and which mammograms cannot, thus needing to be read by a radiologist. Results obtained on a private dataset of 8,162 patients show that MAMMO reduced the number of radiologist readings by 42.8% while improving the overall diagnostic accuracy in comparison to readings done by radiologists alone. We analyze the triage of patients decided by MAMMO to gain a better understanding of what unique mammogram characteristics require radiologists' expertise.

We explore how Deep Learning (DL) can be utilized to predict prognosis of acute myeloid leukemia (AML). Out of TCGA (The Cancer Genome Atlas) database, 94 AML cases are used in this study. Input data include age, 10 common cytogenetic and 23 most common mutation results; output is the prognosis (diagnosis to death, DTD). In our DL network, autoencoders are stacked to form a hierarchical DL model from which raw data are compressed and organized and high-level features are extracted. The network is written in R language and is designed to predict prognosis of AML for a given case (DTD of more than or less than 730 days). The DL network achieves an excellent accuracy of 83% in predicting prognosis. As a proof-of-concept study, our preliminary results demonstrate a practical application of DL in future practice of prognostic prediction using next-gen sequencing (NGS) data.