Deep neural networks (DNNs) have achieved state-of-the-art performance across a variety of traditional machine learning tasks, e.g., speech recognition, image classification, and segmentation. The ability of DNNs to efficiently approximate high-dimensional functions has also motivated their use in scientific applications, e.g., to solve partial differential equations (PDE) and to generate surrogate models. In this paper, we consider the supervised training of DNNs, which arises in many of the above applications. We focus on the central problem of optimizing the weights of the given DNN such that it accurately approximates the relation between observed input and target data. Devising effective solvers for this optimization problem is notoriously challenging due to the large number of weights, non-convexity, data-sparsity, and non-trivial choice of hyperparameters. To solve the optimization problem more efficiently, we propose the use of variable projection (VarPro), a method originally designed for separable nonlinear least-squares problems. Our main contribution is the Gauss-Newton VarPro method (GNvpro) that extends the reach of the VarPro idea to non-quadratic objective functions, most notably, cross-entropy loss functions arising in classification. These extensions make GNvpro applicable to all training problems that involve a DNN whose last layer is an affine mapping, which is common in many state-of-the-art architectures. In numerical experiments from classification and surrogate modeling, GNvpro not only solves the optimization problem more efficiently but also yields DNNs that generalize better than commonly-used optimization schemes.

Traditional methods for solvability region analysis can only have inner approximations with inconclusive conservatism. Machine learning methods have been proposed to approach the real region. In this letter, we propose a deep active learning framework for power system solvability prediction. Compared with the passive learning methods where the training is performed after all instances are labeled, the active learning selects most informative instances to be label and therefore significantly reduce the size of labeled dataset for training. In the active learning framework, the acquisition functions, which correspond to different sampling strategies, are defined in terms of the on-the-fly posterior probability from the classifier. The IEEE 39-bus system is employed to validate the proposed framework, where a two-dimensional case is illustrated to visualize the effectiveness of the sampling method followed by the full-dimensional numerical experiments.

This article is co-written together with Syed Nazmus Sadat who Studies Forestry and Environmental Science at Shahjalal University of Science & Technology, Sylhet in Bangladesh. How can artificial intelligence help in efforts to prevent deforestation? Deforestation has an incredibly adverse impact on planet earth. The forests cover close to a third of the land area on our planet and provide us with purer air and fresher water. Eighty percent of the world's land based wildlife live in forests [1].

"Although scientists have a good understanding how much carbon is in the atmosphere, it's surprisingly tough to trace where those emissions come from," says Gavin McCormick, the founder of a nonprofit called WattTime that also makes technology that enables smart devices to automatically reduce emissions. The startup is working with several other climate and tech organizations and the former vice president Al Gore on the new project. Right now, McCormick says, most emissions data is self-reported, and it can sometimes take years for the data to be gathered. "We think that technology, in particular AI and satellites, have the potential to change that pretty profoundly, which can influence sort of any sector that depends on really knowing where emissions are coming from to make good decisions," he says. "The time lag in current data makes it often non-actionable," says Gore, who has been helping structure the project to have the maximum impact on the climate crisis and enlisting partners for financial and strategic support.

In spatial statistics, a common objective is to predict the values of a spatial process at unobserved locations by exploiting spatial dependence. In geostatistics, Kriging provides the best linear unbiased predictor using covariance functions and is often associated with Gaussian processes. However, when considering non-linear prediction for non-Gaussian and categorical data, the Kriging prediction is not necessarily optimal, and the associated variance is often overly optimistic. We propose to use deep neural networks (DNNs) for spatial prediction. Although DNNs are widely used for general classification and prediction, they have not been studied thoroughly for data with spatial dependence. In this work, we propose a novel neural network structure for spatial prediction by adding an embedding layer of spatial coordinates with basis functions. We show in theory that the proposed DeepKriging method has multiple advantages over Kriging and classical DNNs only with spatial coordinates as features. We also provide density prediction for uncertainty quantification without any distributional assumption and apply the method to PM$_{2.5}$ concentrations across the continental United States.

A cyclist passes electric automobiles charging at Ubeeqo SAS electric vehicle charge stations in ... [ ] Paris, France, on Wednesday, May 27, 2020. President Emmanuel Macron's plan includes incentives for the purchase of electric cars, cash-for-clunkers to encourage consumers to trade in older, more polluting cars and subsidies for struggling car-parts makers. In November 2019, the first case of Covid-19 was reported in Wuhan, China. During the early days of the outbreak, local authorities attempted to clamp down on sharing information about the virus, but as the transmission strengthened in the region, the government imposed lockdown measures across China's Hubei province to control the spread of Covid-19. On January 22, Wuhan became the first major city under quarantine, and in the months that followed, many cities followed suit that caused a shock to the global economy.

Recent progress in scientific machine learning (SciML) has opened up the possibility of training novel neural network architectures that solve complex partial differential equations (PDEs). Several (nearly data free) approaches have been recently reported that successfully solve PDEs, with examples including deep feed forward networks, generative networks, and deep encoder-decoder networks. However, practical adoption of these approaches is limited by the difficulty in training these models, especially to make predictions at large output resolutions ($\geq 1024 \times 1024$). Here we report on a software framework for data parallel distributed deep learning that resolves the twin challenges of training these large SciML models - training in reasonable time as well as distributing the storage requirements. Our framework provides several out of the box functionality including (a) loss integrity independent of number of processes, (b) synchronized batch normalization, and (c) distributed higher-order optimization methods. We show excellent scalability of this framework on both cloud as well as HPC clusters, and report on the interplay between bandwidth, network topology and bare metal vs cloud. We deploy this approach to train generative models of sizes hitherto not possible, showing that neural PDE solvers can be viably trained for practical applications. We also demonstrate that distributed higher-order optimization methods are $2-3\times$ faster than stochastic gradient-based methods and provide minimal convergence drift with higher batch-size.

Based on the last two wildfire seasons, including 2018 when an entire California town was destroyed, utilities blamed for recent wildfires need all the help they can get maintaining aging grids. AI technologies may provide new monitoring tools. Paradise, Calif., population of about 27,000, was destroyed by the Camp Fire. The 2018 inferno claimed at least 84 victims. In June, Pacific Gas & Electric (PG&E) was ordered to pay a $3.5 million fine for causing the Camp Fire.

Recently, Expo 2020 Dubai announced Terminus Group as its Official Robotics Partner. Mohammed Al Hashmi, CTO of Expo 2020 Dubai disclosed that more than 150 smart robots will be deployed at the Expo. For this collaboration, Victor Ai, Founder and CEO of Terminus Group, commented: "We have been committing to the intersection of science, engineering and technology, which empowered Terminus Group's Robots with world-leading interactive computing. Undoubtedly, Terminus Group is a prior choice for any group requiring high-quality robotics product". Together with Opti, the mascot robot designed by Terminus Group exclusively for the purposes of the Expo 2020 Dubai event, the robots will interact with the visitors from all around the world by offering diverse multi-modal interactions, 5G network connectivity, AI mapping and object detection.

Bureaucrats -- particularly those from the European Union (EU) -- rarely get the praise they deserve. By their nature, they are reserved, so they do not draw attention to themselves when things go well. When things go poorly, though, they make for a convenient target. So when the EU does something bold, we should give it its due. The EU's boldness in addressing a host of environmental problems head-on is unmatched.