We empirically characterize the performance of discriminative and generative LSTM models for text classification. We find that although RNN-based generative models are more powerful than their bag-of-words ancestors (e.g., they account for conditional dependencies across words in a document), they have higher asymptotic error rates than discriminatively trained RNN models. However we also find that generative models approach their asymptotic error rate more rapidly than their discriminative counterparts---the same pattern that Ng & Jordan (2001) proved holds for linear classification models that make more naive conditional independence assumptions. Building on this finding, we hypothesize that RNN-based generative classification models will be more robust to shifts in the data distribution. This hypothesis is confirmed in a series of experiments in zero-shot and continual learning settings that show that generative models substantially outperform discriminative models.

Biomedical researchers use ontologies to annotate their data with ontology terms, enabling better data integration and interoperability. However, the number, variety and complexity of current biomedical ontologies make it cumbersome for researchers to determine which ones to reuse for their specific needs. To overcome this problem, in 2010 the National Center for Biomedical Ontology (NCBO) released the Ontology Recommender, which is a service that receives a biomedical text corpus or a list of keywords and suggests ontologies appropriate for referencing the indicated terms. We developed a new version of the NCBO Ontology Recommender. Called Ontology Recommender 2.0, it uses a new recommendation approach that evaluates the relevance of an ontology to biomedical text data according to four criteria: (1) the extent to which the ontology covers the input data; (2) the acceptance of the ontology in the biomedical community; (3) the level of detail of the ontology classes that cover the input data; and (4) the specialization of the ontology to the domain of the input data. Our evaluation shows that the enhanced recommender provides higher quality suggestions than the original approach, providing better coverage of the input data, more detailed information about their concepts, increased specialization for the domain of the input data, and greater acceptance and use in the community. In addition, it provides users with more explanatory information, along with suggestions of not only individual ontologies but also groups of ontologies. It also can be customized to fit the needs of different scenarios. Ontology Recommender 2.0 combines the strengths of its predecessor with a range of adjustments and new features that improve its reliability and usefulness. Ontology Recommender 2.0 recommends over 500 biomedical ontologies from the NCBO BioPortal platform, where it is openly available.

Background-Foreground classification is a fundamental well-studied problem in computer vision. Due to the pixel-wise nature of modeling and processing in the algorithm, it is usually difficult to satisfy real-time constraints. There is a trade-off between the speed (because of model complexity) and accuracy. Inspired by the rejection cascade of Viola-Jones classifier, we decompose the Gaussian Mixture Model (GMM) into an adaptive cascade of classifiers. This way we achieve a good improvement in speed without compensating for accuracy. In the training phase, we learn multiple KDEs for different durations to be used as strong prior distribution and detect probable oscillating pixels which usually results in misclassifications. We propose a confidence measure for the classifier based on temporal consistency and the prior distribution. The confidence measure thus derived is used to adapt the learning rate and the thresholds of the model, to improve accuracy. The confidence measure is also employed to perform temporal and spatial sampling in a principled way. We demonstrate a speed-up factor of 5x to 10x and 17 percent average improvement in accuracy over several standard videos.

The computations of polynomials (over a field, which we shall throughout assume is of zero or large enough characteristic) using arithmetic operations of addition and multiplication (and possibly division) are of course as natural as the computation of Boolean functions via logical gates, and capture many natural important tasks including Fourier transforms, linear algebra, matrix computations and more generally symbolic algebraic computations arising in many settings. Arithmetic circuits are the natural computational model for understanding the computational complexity of such tasks just like Boolean circuits are for Boolean functions. The presence of algebraic structure and mathematical tools supplied by centuries of work in algebra were a source of hope that understanding arithmetic circuits will be much faster and easier than their Boolean siblings. And while we generally know more about arithmetic circuits, their power is far from understood, and in particular, the arithmetic analog VP vs. VNP of the Boolean P vs. NP problem as formulated by Valiant8 is wide open. The past few years have seen a revolution in our understanding of arithmetic circuits.

We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry. Four years ago, a paper by Yann LeCun and his collaborators was rejected by the leading computer vision conference on the grounds that it used neural networks and therefore provided no insight into how to design a vision system. At the time, most computer vision researchers believed that a vision system needed to be carefully hand-designed using a detailed understanding of the nature of the task. They assumed that the task of classifying objects in natural images would never be solved by simply presenting examples of images and the names of the objects they contained to a neural network that acquired all of its knowledge from this training data. What many in the vision research community failed to appreciate was that methods that require careful hand-engineering by a programmer who understands the domain do not scale as well as methods that replace the programmer with a powerful general-purpose learning procedure.

Since its inauguration in 1966, the ACM A.M. Turing Award has recognized major contributions of lasting importance to computing. Through the years, it has become the most prestigious award in computing. To help celebrate 50 years of the ACM Turing Award and the visionaries who have received it, ACM has launched a campaign called "Panels in Print," which takes the form of a collection of responses from Turing laureates, ACM award recipients and other ACM experts on a given topic or trend. For our fourth and final Panel in Print, we invited 2014 ACM A.M. Turing Award recipient MICHAEL STONEBRAKER, 2013 ACM Prize recipient DAVID BLEI, 2007 ACM Prize recipient DAPHNE KOLLER, and ACM Fellow VIPIN KUMAR to discuss trends in big data. Gartner estimates that there are currently about 4.9 billion connected devices (cars, homes, appliances, industrial equipment, among others) generating data.

NASA's Exploration Mission-1 has stepped closer to reality, as the space agency completed the second flight controller tests for the engines that will power its'megarocket.' Engineers conducted a 500-second test on the component said to be the'brain' of the RS-25 engines – the four engines that will simultaneously provide 2 million pounds of thrust for the Space Launch System (SLS). NASA previously tested the first flight controller in March ahead of installation in one of the EM-1 engines, and once they've reviewed the new data, the second controller will be installed. NASA's Exploration Mission-1 has stepped closer to reality, as the space agency completed the second flight controller tests for the engines that will power its'megarocket.' Engineers conducted a 500-second test on the component said to be the'brain' of the RS-25 engines Nasa's Orion, stacked on a Space Launch System rocket capable of lifting 70 metric tons will launch from a newly refurbished Kennedy Space Center in 2019.

Machine learning (ML) has received a lot of attention recently, and not without good reason. It has already revolutionized fields from image recognition to healthcare to transportation.

The forecast is frightening: Robots will take over all manual labor and self-generating code will automatically spin out the algorithms once developed by statisticians and programmers. What will mere mortals do all day long? Ride captive as our self-driving cars take us on a sentimental journey to see the parking lots where shopping malls used to be? Visit the local greenhouse to watch through the window as mechanical gardeners harvest hydroponic vegetables loaded onto drones optimized for doorstep delivery? Or just sit in front of a giant screen forever and ever while Amaflix Consolidated serves up an endless number of Arrested Development re-runs because AmaFlix knows that, no matter what clever new show we might tell our friends we are binge-watching, we are really just watching Arrested Development re-runs over and over again. Suffice to say: The standard error of Y estimates for X-axis values outside the range of observed data tend to be large for a reason.

With major technology companies and startups seriously embracing Cloud strategies, now is the perfect time to attend @CloudExpo @ThingsExpo, June 6-8, 2017, at the Javits Center in New York City, NY and October 31 - November 2, 2017, Santa Clara Convention Center, CA. Join Cloud Expo / @ThingsExpo conference chair Roger Strukhoff (@IoT2040), June 6-8, 2017, at the Javits Center in New York City, NY and October 31 - November 2, 2017, Santa Clara Convention Center, CA for three days of intense Enterprise Cloud and'Digital Transformation' discussion and focus, including Big Data's indispensable role in IoT, Smart Grids and (IIoT) Industrial Internet of Things, Wearables and Consumer IoT, as well as (new) Digital Transformation in Vertical Markets. Accordingly, attendees at the upcoming 20th Cloud Expo / @ThingsExpo June 6-8, 2017, at the Javits Center in New York City, NY and October 31 - November 2, 2017, Santa Clara Convention Center, CA will find fresh new content in a new track called FinTech, which will incorporate machine learning, artificial intelligence, deep learning, and blockchain into one track. The upcoming 20th International @CloudExpo @ThingsExpo, June 6-8, 2017, at the Javits Center in New York City, NY and October 31 - November 2, 2017, Santa Clara Convention Center, CA announces that its Call For Papers for speaking opportunities is open.