Over time, TensorFlow has grown in popularity and is now being used by developers for solving problems using deep learning methods for image recognition, video detection, text processing like sentiment analysis, etc. Like any other library, you may need some time to get used to the concepts that TensorFlow is built on. And, once you do, with the help of documentation and community support, representing problems as data graphs and solving them with TensorFlow can make machine learning at scale a less tedious process. In TensorFlow, constants are created using the constant function which takes a few parameters: Value, dtype (data type), shape, name and (verify_shape) shape verification.

Anyone who has thought about scaling a business or building a network is familiar with a dynamic referred to as the "network effect." The more buyers and sellers who use a marketplace like eBay, for example, the more useful it becomes. Well, the data network effect is a dynamic in which increased use of a service actually improves the service, such as how machine-learning models generally grow more accurate as a result of training from larger and larger volumes of data. Autonomous vehicles and other smart robots rely on sensors that generate increasingly massive volumes of highly varied data. This data is used to build better AI models that robots rely on to make real-time decisions and navigate real-world environments.

Deep learning, an advanced machine-learning technique, uses layered (hence "deep") neural networks (neural nets) that are loosely modelled on the human brain. Machine learning itself is a subset of Artificial Intelligence (AI), and is broadly about teaching a computer how to spot patterns and use mountains of data to make connections without any programming to accomplish the specific task--a recommendation engine being a good example. Neural nets, on their part, enable image recognition, speech recognition, self-driving cars and smarthome automation devices, among other things. However, the success of deep learning is primarily dependent on the availability of huge data sets on which these neural nets can be trained, coupled with a lot of computing power, memory and energy to function. To address this issue, says a 14 November press release, researchers at the University of Waterloo, Canada, took a cue from nature to make this process more efficient, thus making deep-learning software compact enough to fit on mobile computer chips for use in everything from smartphones to industrial robots.

This summer, Intel has been collaborating with the NASA Frontier Development Lab (FDL), an AI R&D accelerator targeting knowledge gaps useful to the space program. The NASA FDL, hosted at the SETI Institute, was established to apply AI to five specific challenges in areas relevant to the space program: Planetary Defense (defending the Earth from potentially hazardous asteroids), Space Weather (better predicting solar activity) and Space Resources (locating and accessing the resources we'll need to go back to the moon and expand into the solar system). Earlier this summer, we introduced you to this collaboration, and we have exciting updates to share. The NASA FDL team successfully applied the Intel Nervana Deep Learning platform to automate the creation of lunar maps at our Moon's poles – a critical step in helping both identify potential landing sites and navigation in the shadowed regions of the moon. Here, permanent darkness and extremely low temperatures make for an ideal location for water ice (and other volatiles), but highly challenging conditions for future missions that would be impossible without detailed mapping.

We propose a framework that learns a representation transferable across different domains and tasks in a label efficient manner. Our approach battles domain shift with a domain adversarial loss, and generalizes the embedding to novel task using a metric learning-based approach. Our model is simultaneously optimized on labeled source data and unlabeled or sparsely labeled data in the target domain. Our method shows compelling results on novel classes within a new domain even when only a few labeled examples per class are available, outperforming the prevalent fine-tuning approach. In addition, we demonstrate the effectiveness of our framework on the transfer learning task from image object recognition to video action recognition.

We explore whether useful temporal neural generative models can be learned from sequential data without back-propagation through time. We investigate the viability of a more neurocognitively-grounded approach in the context of unsupervised generative modeling of sequences. Specifically, we build on the concept of predictive coding, which has gained influence in cognitive science, in a neural framework. To do so we develop a novel architecture, the Temporal Neural Coding Network, and its learning algorithm, Discrepancy Reduction. The underlying directed generative model is fully recurrent, meaning that it employs structural feedback connections and temporal feedback connections, yielding information propagation cycles that create local learning signals. This facilitates a unified bottom-up and top-down approach for information transfer inside the architecture. Our proposed algorithm shows promise on the bouncing balls generative modeling problem. Further experiments could be conducted to explore the strengths and weaknesses of our approach.

In this paper, we show that the recent integration of statistical models with deep recurrent neural networks provides a new way of formulating volatility (the degree of variation of time series) models that have been widely used in time series analysis and prediction in finance. The model comprises a pair of complementary stochastic recurrent neural networks: the generative network models the joint distribution of the stochastic volatility process; the inference network approximates the conditional distribution of the latent variables given the observables. Our focus here is on the formulation of temporal dynamics of volatility over time under a stochastic recurrent neural network framework. Experiments on real-world stock price datasets demonstrate that the proposed model generates a better volatility estimation and prediction that outperforms stronge baseline methods, including the deterministic models, such as GARCH and its variants, and the stochastic MCMC-based models, and the Gaussian-process-based, on the average negative log-likelihood measure.

A low-cost, robust, and simple mechanism to measure hemoglobin would play a critical role in the modern health infrastructure. Consistent sample acquisition has been a long-standing technical hurdle for photometer-based portable hemoglobin detectors which rely on micro cuvettes and dry chemistry. Any particulates (e.g. intact red blood cells (RBCs), microbubbles, etc.) in a cuvette's sensing area drastically impact optical absorption profile, and commercial hemoglobinometers lack the ability to automatically detect faulty samples. We present the ground-up development of a portable, low-cost and open platform with equivalent accuracy to medical-grade devices, with the addition of CNN-based image processing for rapid sample viability prechecks. The developed platform has demonstrated precision to the nearest $0.18[g/dL]$ of hemoglobin, an R^2 = 0.945 correlation to hemoglobin absorption curves reported in literature, and a 97% detection accuracy of poorly-prepared samples. We see the developed hemoglobin device/ML platform having massive implications in rural medicine, and consider it an excellent springboard for robust deep learning optical spectroscopy: a currently untapped source of data for detection of countless analytes.

With the development of segmentation and classification methods of 3D point clouds by machine-learning, more and more data are needed in quantity and quality (number of points, number of classes, quality of segmentation). 1 Figure 1: Part of our dataset (top: reflectance from blue(0) to red(255), middle: object label (different color for each), bottom: object class) 2 There are always more datasets of classification and segmentation of images, visual and LiDAR odometry or SLAM, detection of vehicles and pedestrians on videos, stereo-vision, optical flow, etc. But it is still difficult to find datasets of segmented and classified urban 3D point clouds. The only comparable datasets are the one described in section Available Datasets. Each of them have their advantages and disadvantages, but we estimate that none has the quality and quantity required for new issues such as deep learning methods. In section Our Dataset: Paris-Lille-3D, we present a new urban dataset that we have created, where the objects are sufficiently segmented that the task of segmentation can be learned very precisely. Our dataset can be found at the following address: http://caor-mines-paristech.fr/fr/paris-lille-

Predicting highrisk vascular diseases is a significant issue in the medical domain. Most predicting methods predict the prognosis of patients from pathological and radiological measurements, which are expensive and require much time to be analyzed. Here we propose deep attention models that predict the onset of the high risky vascular disease from symbolic medical histories sequence of hypertension patients such as ICD-10 and pharmacy codes only, Medical History-based Prediction using Attention Network (MeHPAN). We demonstrate two types of attention models based on 1) bidirectional gated recurrent unit (R-MeHPAN) and 2) 1D convolutional multilayer model (C-MeHPAN). Two MeHPAN models are evaluated on approximately 50,000 hypertension patients with respect to precision, recall, f1-measure and area under the curve (AUC). Experimental results show that our MeHPAN methods outperform standard classification models. Comparing two MeHPANs, R-MeHPAN provides more better discriminative capability with respect to all metrics while C-MeHPAN presents much shorter training time with competitive accuracy.