In the last decades wind power became the second largest energy source in the EU covering 16% of its electricity demand. However, due to its volatility, accurate short range wind power predictions are required for successful integration of wind energy into the electrical grid. Accurate predictions of wind power require accurate hub height wind speed forecasts, where the state of the art method is the probabilistic approach based on ensemble forecasts obtained from multiple runs of numerical weather prediction models. Nonetheless, ensemble forecasts are often uncalibrated and might also be biased, thus require some form of post-processing to improve their predictive performance. We propose a novel flexible machine learning approach for calibrating wind speed ensemble forecasts, which results in a truncated normal predictive distribution. In a case study based on 100m wind speed forecasts produced by the operational ensemble prediction system of the Hungarian Meteorological Service, the forecast skill of this method is compared with the predictive performance of three different ensemble model output statistics approaches and the raw ensemble forecasts. We show that compared with the raw ensemble, post-processing always improves the calibration of probabilistic and accuracy of point forecasts and from the four competing methods the novel machine learning based approach results in the best overall performance.

We present PyTorch Geometric Temporal a deep learning framework combining state-of-the-art machine learning algorithms for neural spatiotemporal signal processing. The main goal of the library is to make temporal geometric deep learning available for researchers and machine learning practitioners in a unified easy-to-use framework. PyTorch Geometric Temporal was created with foundations on existing libraries in the PyTorch eco-system, streamlined neural network layer definitions, temporal snapshot generators for batching, and integrated benchmark datasets. These features are illustrated with a tutorial-like case study. Experiments demonstrate the predictive performance of the models implemented in the library on real world problems such as epidemiological forecasting, ridehail demand prediction and web-traffic management. Our sensitivity analysis of runtime shows that the framework can potentially operate on web-scale datasets with rich temporal features and spatial structure.

Google this week pushed back against claims by earlier research that large AI models can contribute significantly to carbon emissions. In a paper coauthored by Google AI chief scientist Jeff Dean, researchers at the company say that the choice of model, datacenter, and processor can reduce carbon footprint by up to 100 times and that "misunderstandings" about the model lifecycle contributed to "miscalculations" in impact estimates. Carbon dioxide, methane, and nitrous oxide levels are at the highest they've been in the last 800,000 years. Together with other drivers, greenhouse gases likely catalyzed the global warming that's been observed since the mid-20th century. It's widely believed that machine learning models, too, have contributed to the adverse environmental trend.

Climate change is the planet's greatest challenge. The UN has already stated that 2021 is the final year for us to make real change in the fight against rising global temperatures. The UN organization is hosting the COP26 climate summit to address this dilemma of the century, where major players like Hitachi and BCG are involved as partners in this critical effort. Moreover, with Climate AI Champions in the picture, these innovators could provide the right solutions we need in the fight for survival and growth. The climate change crisis is real, finding quick and affordable solutions is an urgency, and AI can play a major role.

How can I prepare myself for something I do not yet know? Scientists from the Technical University of Munich and from the Fritz Haber Institute in Berlin have addressed this almost philosophical question in the context of machine learning. Learning is no more than drawing new decisions on prior experience. In order to deal with a new situation in this way, one needs to have dealt with roughly similar situations before. In machine learning, this correspondingly means that a learning algorithm needs to have been exposed to roughly similar data.

We consider the distributed training of large-scale neural networks that serve as PDE solvers producing full field outputs. We specifically consider neural solvers for the generalized 3D Poisson equation over megavoxel domains. A scalable framework is presented that integrates two distinct advances. First, we accelerate training a large model via a method analogous to the multigrid technique used in numerical linear algebra. Here, the network is trained using a hierarchy of increasing resolution inputs in sequence, analogous to the 'V', 'W', 'F', and 'Half-V' cycles used in multigrid approaches. In conjunction with the multi-grid approach, we implement a distributed deep learning framework which significantly reduces the time to solve. We show the scalability of this approach on both GPU (Azure VMs on Cloud) and CPU clusters (PSC Bridges2). This approach is deployed to train a generalized 3D Poisson solver that scales well to predict output full-field solutions up to the resolution of 512x512x512 for a high dimensional family of inputs.

Gaussian processes (GPs) are non-linear probabilistic models popular in many applications. However, na\"ive GP realizations require quadratic memory to store the covariance matrix and cubic computation to perform inference or evaluate the likelihood function. These bottlenecks have driven much investment in the development of approximate GP alternatives that scale to the large data sizes common in modern data-driven applications. We present in this manuscript MuyGPs, a novel efficient GP hyperparameter estimation method. MuyGPs builds upon prior methods that take advantage of the nearest neighbors structure of the data, and uses leave-one-out cross-validation to optimize covariance (kernel) hyperparameters without realizing a possibly expensive likelihood. We describe our model and methods in detail, and compare our implementations against the state-of-the-art competitors in a benchmark spatial statistics problem. We show that our method outperforms all known competitors both in terms of time-to-solution and the root mean squared error of the predictions.

Graph representation learning has become a ubiquitous component in many scenarios, ranging from social network analysis to energy forecasting in smart grids. In several applications, ensuring the fairness of the node (or graph) representations with respect to some protected attributes is crucial for their correct deployment. Yet, fairness in graph deep learning remains under-explored, with few solutions available. In particular, the tendency of similar nodes to cluster on several real-world graphs (i.e., homophily) can dramatically worsen the fairness of these procedures. In this paper, we propose a biased edge dropout algorithm (FairDrop) to counter-act homophily and improve fairness in graph representation learning. FairDrop can be plugged in easily on many existing algorithms, is efficient, adaptable, and can be combined with other fairness-inducing solutions. After describing the general algorithm, we demonstrate its application on two benchmark tasks, specifically, as a random walk model for producing node embeddings, and to a graph convolutional network for link prediction. We prove that the proposed algorithm can successfully improve the fairness of all models up to a small or negligible drop in accuracy, and compares favourably with existing state-of-the-art solutions. In an ablation study, we demonstrate that our algorithm can flexibly interpolate between biasing towards fairness and an unbiased edge dropout. Furthermore, to better evaluate the gains, we propose a new dyadic group definition to measure the bias of a link prediction task when paired with group-based fairness metrics. In particular, we extend the metric used to measure the bias in the node embeddings to take into account the graph structure.

The Encyclopedia Britannica defines Artificial Intelligence or A.I. as "the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings." Intelligent beings are those who can adapt to changing circumstances. The most forward-thinking companies are investing in Artificial Intelligence, as they already realized the importance of A.I. in business, and the impact A.I. will have, while it is becoming a key component of organizations' strategies as digital disruption increases. I am sharing here today an overview of the top 5 companies in the world according to Fortune 2020) and some examples of how these companies are using A.I. to empower their business. Walmart has been in business since the 1960s, but the company is still developing ways to revolutionize retail operations and enhance customer service.

Preface: Though CosmiQ Works (and its associated blog: The DownLinQ) has unfortunately been shut down, there remains much to be done in the geospatial analytics domain. Accordingly, this blog details work performed independently of IQT and in my spare time. In a number of previous blogs [e.g. 1, 2, 3] and academic papers [e.g. 4, 5, 6] we've demonstrated the striking efficacy of adapting YOLO to detect objects in satellite imagery. Recall that YOLO is a leading deep learning object detection framework, designed to detect objects in imagery. YOLO maxes out at image sizes of a few thousand pixels in size, far too small to handle large scale satellite imagery which can exceed 100 million pixels.