The vulnerability of deep networks to adversarial attacks is a central problem for deep learning from the perspective of both cognition and security. The current most successful defense method is to train a classifier using adversarial images created during learning. Another defense approach involves transformation or purification of the original input to remove adversarial signals before the image is classified. We focus on defending naturally-trained classifiers using Markov Chain Monte Carlo (MCMC) sampling with an Energy-Based Model (EBM) for adversarial purification. In contrast to adversarial training, our approach is intended to secure pre-existing and highly vulnerable classifiers. The memoryless behavior of long-run MCMC sampling will eventually remove adversarial signals, while metastable behavior preserves consistent appearance of MCMC samples after many steps to allow accurate long-run prediction. Balancing these factors can lead to effective purification and robust classification. We evaluate adversarial defense with an EBM using the strongest known attacks against purification. Our contributions are 1) an improved method for training EBM's with realistic long-run MCMC samples, 2) an Expectation-Over-Transformation (EOT) defense that resolves theoretical ambiguities for stochastic defenses and from which the EOT attack naturally follows, and 3) state-of-the-art adversarial defense for naturally-trained classifiers and competitive defense compared to adversarially-trained classifiers on Cifar-10, SVHN, and Cifar-100. Code and pre-trained models are available at https://github.com/point0bar1/ebm-defense.

Street lamps constitute the densest electrically operated public infrastructure in urban areas. Their changeover to energy-friendly LED light quickly amortizes and is increasingly leveraged for smart city projects, where LED street lamps double, for example, as wireless networking or sensor infrastructure. We make the case for a new paradigm called SLaaP--street lamps as a platform. SLaaP is proposed as an open, enabling platform, fostering innovative citywide services for the full range of stakeholders and end users--seamlessly extending from everyday use to emergency response. In this article, we first describe the role and potential of street lamps and introduce one novel base service as a running example. We then discuss citywide infrastructure design and operation, followed by addressing the major layers of a SLaaP infrastructure: hardware, distributed software platform, base services, value-added services and applications for users and'things.' Finally, we discuss the crucial roles and participation of major stakeholders: citizens, city, government, and economy. Recent years have seen the emergence of smart street lamps, with very different meanings of'smart'--sometimes related to the original purpose as with usage-dependent lighting, but mostly as add-on capabilities like urban sensing, monitoring, digital signage, WiFi access, or e-vehicle charging.a The future holds even more use cases: for example, after a first wave of 5G mobile network rollouts from 2020 onward, a second wave shall apply mm-wave frequencies for which densely deployed light poles can be appropriate'cell towers.'

"This is a bill being sold as a privacy bill, but it's a wolf in sheep's clothing," Matt Cagle, an attorney for the American Civil Liberties Union of Northern California, said in an interview. The ACLU, Electronic Frontier Foundation and other civil liberties groups held a virtual rally Thursday night to rail against the bill, calling it vaguely worded and potentially dangerous for low-income communities hit hard by the coronavirus. Their remarks were the latest shots fired from a campaign to halt the legislation. The bill's fate in California--which has pushed for more aggressive privacy protections in recent years--could foreshadow how a potentially huge market for facial recognition technology is regulated by other states. The bill calls for companies and agencies that use facial recognition tools in areas accessible to the public to "provide a conspicuous and contextually appropriate notice" that faces may get scanned.

Safe autonomous navigation is an essential and challenging problem for robots operating in highly unstructured or completely unknown environments. Under these conditions, not only robotic systems must deal with limited localisation information, but also their manoeuvrability is constrained by their dynamics and often suffer from uncertainty. In order to cope with these constraints, this manuscript proposes an uncertainty-based framework for mapping and planning feasible motions online with probabilistic safety-guarantees. The proposed approach deals with the motion, probabilistic safety, and online computation constraints by: (i) incrementally mapping the surroundings to build an uncertainty-aware representation of the environment, and (ii) iteratively (re)planning trajectories to goal that are kinodynamically feasible and probabilistically safe through a multi-layered sampling-based planner in the belief space. In-depth empirical analyses illustrate some important properties of this approach, namely, (a) the multi-layered planning strategy enables rapid exploration of the high-dimensional belief space while preserving asymptotic optimality and completeness guarantees, and (b) the proposed routine for probabilistic collision checking results in tighter probability bounds in comparison to other uncertainty-aware planners in the literature. Furthermore, real-world in-water experimental evaluation on a non-holonomic torpedo-shaped autonomous underwater vehicle and simulated trials in the Stairwell scenario of the DARPA Subterranean Challenge 2019 on a quadrotor unmanned aerial vehicle demonstrate the efficacy of the method as well as its suitability for systems with limited on-board computational power.

Recently, a groundswell of research has identified the use of counterfactual explanations as a potentially significant solution to the Explainable AI (XAI) problem. It is argued that (a) technically, these counterfactual cases can be generated by permuting problem-features until a class change is found, (b) psychologically, they are much more causally informative than factual explanations, (c) legally, they are GDPR-compliant. However, there are issues around the finding of good counterfactuals using current techniques (e.g. sparsity and plausibility). We show that many commonly-used datasets appear to have few good counterfactuals for explanation purposes. So, we propose a new case based approach for generating counterfactuals using novel ideas about the counterfactual potential and explanatory coverage of a case-base. The new technique reuses patterns of good counterfactuals, present in a case-base, to generate analogous counterfactuals that can explain new problems and their solutions. Several experiments show how this technique can improve the counterfactual potential and explanatory coverage of case-bases that were previously found wanting.

Recently, machine learning tools in particular neural networks have been widely used to solve differential equations. One main advantage of using machine learning, in this case, is that one does not need to mesh the computational domain and can instead randomly draw data points to solve the differential equations of interest. In this work, we propose a simple neural network to approximate low-frequency periodic functions or seek such solutions of differential equations. To this end, we build a Fourier Neural Network (FNN) represented as a shallow neural network (i.e with one hidden layer) based on the Fourier Decomposition. As opposed to traditional neural networks, which feature activation functions such as the sigmoid, logistic, ReLU, hyperbolic tangent and softmax functions, Fourier Neural Networks are composed using sinusoidal activation functions. We propose a strategy to initialize the weights of this FNN and showcase its performance against traditional networks for function approximations and differential equations solutions.

Wildfire modelling is an attempt to reproduce fire behaviour. Through active fire analysis, it is possible to reproduce a dynamical process, such as wildfires, with limited duration time series data. Recurrent neural networks (RNNs) can model dynamic temporal behaviour due to their ability to remember their internal input. In this paper, we compare the Gated Recurrent Unit (GRU) and the Long Short-Term Memory (LSTM) network. We try to determine whether a wildfire continues to burn and given that it does, we aim to predict which one of the 8 cardinal directions the wildfire will spread in. Overall the GRU performs better for longer time series than the LSTM. We have shown that although we are reasonable at predicting the direction in which the wildfire will spread, we are not able to asses if the wildfire continues to burn due to the lack of auxiliary data.

But 2019 was the year the earth burned. In Australia, the world watched in horror as bushfires destroyed 10.3 million hectares, marking the continent's most intense and destructive fire season in over 40 years. Earlier that fall, California saw more than 101,000 hectares destroyed, with damages upward of $80 billion. Alaska saw nearly a million. Record-breaking fires also hit Indonesia, Russia, Lebanon -- but nowhere saw the sheer mass of media coverage as the fires that tore through the Amazon nearly all last summer. By year's end, thousands of global media outlets had reported that Brazil's largest rainforest played host to more than 80,000 individual forest fires in 2019, resulting in an estimated 906,000 square hectares of environmental destruction. At the time, Brazil's National Institute for Space Research reported it was the fastest rate of burning since record keeping began in 2013. But amid the charred ruins of one of the largest oxygen-producing environments on the planet, a secret lies buried beneath the soil.

The U.S. Air Force is finally pushing into the world of robot combat drones, vowing to fly the first of its "Skyborg" drones by 2023. The service envisions Skyborg as a merging of artificial intelligence with jet-powered drones. The result will be drones capable of flying alongside fighter jets, carrying out dangerous missions. Skyborg drones will be much cheaper than piloted aircraft, allowing the Air Force to grow its fleet at a lower cost. The Air Force, according to Defense News, will award a total of $400 million to one or more companies to develop different types of Skyborg drones.

Councils in the UK expect to save over £195million (€221 million) in 2020 by introducing artificial intelligence technology techniques, according to a national survey of local authorities. Financial savings, faster resolution of enquiries, freeing up staff to focus on citizen engagement and more accurate processing are the four key reasons behind the trend, revealed in a survey of unitary, borough, county and district councils carried out by local government AI and chatbot specialists Agile Datum . Councils each expect to save an average of £300,000 (€340926) in the next 12 months through greater use of artificial intelligence and another £180,000 ( €204556), on average, through the deployment of self-learning chatbots. One in six councils are anticipating savings between £750,000 (€85231 million) and £1m (1.1 million) just around the introduction of artificial intelligence technology. In all, it amounts to savings of £195m (€221 million) across unitary, borough, district and county councils in the UK.