Urbint, a startup developing AI-powered solutions for infrastructure and utility safety, today announced it has raised a $20 million round. The company will use the capital to scale products and expand into new markets and countries, with the goal of delivering "quantifiable" improvements in worker well-being. In the early days of the COVID-19 pandemic, energy companies around the U.S. were forced to send workers home to reduce the spread of infection. This exacerbated many of the industry's longstanding challenges, like how to minimize risks from severe weather, aging infrastructure, and workforce turnover while identifying threats that could cause high-consequence outages to facilities like hospitals and nursing homes. Urbint taps AI to anticipate and prevent catastrophic power failures, with models of the world and machine learning that enable risk-driven decision-making.

Vector autoregressive (VAR) models assume linearity between the endogenous variables and their lags. This linearity assumption might be overly restrictive and could have a deleterious impact on forecasting accuracy. As a solution, we propose combining VAR with Bayesian additive regression tree (BART) models. The resulting Bayesian additive vector autoregressive tree (BAVART) model is capable of capturing arbitrary non-linear relations between the endogenous variables and the covariates without much input from the researcher. Since controlling for heteroscedasticity is key for producing precise density forecasts, our model allows for stochastic volatility in the errors. Using synthetic and real data, we demonstrate the advantages of our methods. For Eurozone data, we show that our nonparametric approach improves upon commonly used forecasting models and that it produces impulse responses to an uncertainty shock that are consistent with established findings in the literature.

In this work, we propose a Bayesian type sparse deep learning algorithm. The algorithm utilizes a set of spike-and-slab priors for the parameters in the deep neural network. The hierarchical Bayesian mixture will be trained using an adaptive empirical method. That is, one will alternatively sample from the posterior using preconditioned stochastic gradient Langevin Dynamics (PSGLD), and optimize the latent variables via stochastic approximation. The sparsity of the network is achieved while optimizing the hyperparameters with adaptive searching and penalizing. A popular SG-MCMC approach is Stochastic gradient Langevin dynamics (SGLD). However, considering the complex geometry in the model parameter space in non-convex learning, updating parameters using a universal step size in each component as in SGLD may cause slow mixing. To address this issue, we apply a computationally manageable preconditioner in the updating rule, which provides a step-size parameter to adapt to local geometric properties. Moreover, by smoothly optimizing the hyperparameter in the preconditioning matrix, our proposed algorithm ensures a decreasing bias, which is introduced by ignoring the correction term in preconditioned SGLD. According to the existing theoretical framework, we show that the proposed algorithm can asymptotically converge to the correct distribution with a controllable bias under mild conditions. Numerical tests are performed on both synthetic regression problems and learning the solutions of elliptic PDE, which demonstrate the accuracy and efficiency of present work.

Learning-to-learn (using optimization algorithms to learn a new optimizer) has successfully trained efficient optimizers in practice. This approach relies on meta-gradient descent on a meta-objective based on the trajectory that the optimizer generates. However, there were few theoretical guarantees on how to avoid meta-gradient explosion/vanishing problems, or how to train an optimizer with good generalization performance. In this paper, we study the learning-to-learn approach on a simple problem of tuning the step size for quadratic loss. Our results show that although there is a way to design the meta-objective so that the meta-gradient remain polynomially bounded, computing the meta-gradient directly using backpropagation leads to numerical issues that look similar to gradient explosion/vanishing problems. We also characterize when it is necessary to compute the meta-objective on a separate validation set instead of the original training set. Finally, we verify our results empirically and show that a similar phenomenon appears even for more complicated learned optimizers parametrized by neural networks.

This study exploits the advancements in information and communication technology (ICT), connected and automated vehicles (CAVs), and sensing, to develop anticipatory multi-objective eco-routing strategies. For a robust application, several GHG costing approaches are examined. The predictive models for the link level traffic and emission states are developed using long short term memory deep network with exogenous predictors. It is found that anticipatory routing strategies outperformed the myopic strategies, regardless of the routing objective. Whether myopic or anticipatory, the multi-objective routing, with travel time and GHG minimization as objectives, outperformed the single objective routing strategies, causing a reduction in the average travel time (TT), average vehicle kilometre travelled (VKT), total GHG and total NOx by 17%, 21%, 18%, and 20%, respectively. Finally, the additional TT and VKT experienced by the vehicles in the network contributed adversely to the amount of GHG and NOx produced in the network.

Bottom Line: Real-time analysis of remote video feeds is rapidly improving thanks to AI, increasing the accuracy of remote equipment and facility monitoring. Agriculture, construction, oil & gas, utilities, and critical infrastructure all need to merge cybersecurity and physical security to adapt to an increasingly complex threatscape. What needs to be the top priority is improving the accuracy, insight, and speed of response to remote threats that AI-based video recognition systems provide. Machine learning techniques as part of a broader AI strategy are proving effective in identifying anomalies and threats in real-time using video, often correlating them back to cyber threats, which are often part of an orchestrated attack on remote facilities. The future of remote security monitoring is being defined by the rapid advances in supervised, unsupervised, and reinforcement machine learning algorithms and their contributions to AI-based visual recognition systems.

In a world of rapid change, artificial intelligence (AI) currently fuels most of this growth so it is no surprise that the next wave of great startups is based in AI solutions. With events like Covid-19, there is increased focus on solutions that tap into the extraordinary capabilities from AI. Something else unique is also happening. Generation Z (Gen Z) and the young millennial entrepreneurs are leading the way. Where are these innovators starting? One area where they are tapping into is a field everyone has had high hopes over the past few decades: robotics.

This Best Artificial Intelligence Stocks forecast is designed for investors and analysts who need predictions for the best companies which are in the frontier of AI application in their products and services. Package Name: Best AI Stocks Recommended Positions: Long Forecast Length: 1 Year (6/23/2019 – 6/24/2020) I Know First Average: 75.08% TSLA was the top performing prediction with a return of 333.09%. The package had an overall average return of 75.08%, providing investors with a premium of 71.70% over the S&P 500's return of 3.38% during the same period. Tesla, Inc., formerly Tesla Motors, Inc., incorporated on July 1, 2003, designs, develops, manufactures and sells fully electric vehicles, and energy storage systems, as well as installs, operates and maintains solar and energy storage products. The Company operates through two segments: automotive, and energy generation and storage.

Images from both the cathode and anode samples which show real and algorithm-generated microstructures. Imperial researchers have demonstrated how machine learning could help design lithium-ion batteries and fuel cells with better performance. A new machine learning algorithm allows researchers to explore possible designs for the microstructure of fuel cells and lithium-ion batteries, before running 3D simulations that help researchers make changes to improve performance. Improvements could include making smartphones charge faster, increasing the time between charges for electric vehicles, and increasing the power of hydrogen fuel cells running data centres. The paper is published in npj Computational Materials.

Almost three years ago, Bill Gates wrote an open letter to the graduating class of 2017 and it's still very relevant today. He offered advice for those looking to make a big impact in the world, and three fields they should consider: artificial intelligence, energy and biosciences. We have only begun to tap into all the ways it will make people's lives more productive and creative. The second is energy, because making it clean, affordable, and reliable will be essential for fighting poverty and climate change. The third is the biosciences, which are ripe with opportunities to help people live longer, healthier lives."