Reinforcement learning algorithms, though successful, tend to over-fit to training environments hampering their application to the real-world. This paper proposes WR$^{2}$L; a robust reinforcement learning algorithm with significant robust performance on low and high-dimensional control tasks. Our method formalises robust reinforcement learning as a novel min-max game with a Wasserstein constraint for a correct and convergent solver. Apart from the formulation, we also propose an efficient and scalable solver following a novel zero-order optimisation method that we believe can be useful to numerical optimisation in general. We contribute both theoretically and empirically. On the theory side, we prove that WR$^{2}$L converges to a stationary point in the general setting of continuous state and action spaces. Empirically, we demonstrate significant gains compared to standard and robust state-of-the-art algorithms on high-dimensional MuJuCo environments.

Incredible footage released by NASA has revealed the space agency's attempts to push its Orion spacecraft's engines to their limits, ahead of a planned 2024 manned mission to the moon dubbed Artemis. In the latest of an on-going series of tests, engineers conducted a continuous 12-minute firing of Orion's propulsion system. Orion is a capsule designed to carry humans to the moon and bring them back safely and the test simulated an abort-to-orbit scenario, in which the second stage of NASA's Space Launch System (SLS) rocket fails. Maggot leaps itself into the air'to catapult to safety' Samsung unveils Galaxy Note10's S Pen that offers greater control Huawei unveils its new'Harmony' phone operating system'Choose truth over facts!' Biden flubs line in Iowa speech Incredible footage released by NASA has revealed the space agency's attempts to push its Orion spacecraft's engines to their limits (pictured), ahead of a planned 2024 manned mission to the moon dubbed Artemis Under ideal conditions the SLS rocket would blast the Orion spacecraft - which will carry astronauts and their supplies - into orbit around the moon. Part of this process involves the interim cryogenic propulsion stage (ICPS) firing, blasting the Orion capsule away from the rocket behind it.

Amazon customers can now adjust the speed of Alexa's voice to make the virtual assistant speak quicker or slower. Beginning today, customers in the US can ask the voice-activated smart speaker to converse slower or faster, depending on their needs. Bosses at the online company have created seven different settings – the standard speaking rate, four faster options and two slower modes. Users can simply simply say'Alexa, speak slower,' or'Alexa, speak faster' to adjust these. They can also reset it by asking it to respond at the default rate.

A computer could be better than a doctor at diagnosing certain types of cancerous and precancerous breast lesions, new research suggests. Researchers at the University of California, Los Angeles, trained an artificial intelligence system using 240 biopsy images, and tested it against 87 pathologists. The machine performed more or less as well as doctors at detecting and classifying all of the breast biopsies. However, it was better at making one crucial distinction: telling the difference between DCIS (ductal carcinoma in situ), a type of cancer, and atypical hyperplasia, a high-risk lesion that has very similar hallmarks but does is not cancerous and does not require the same level of treatment. 'Medical images of breast biopsies contain a great deal of complex data and interpreting them can be very subjective,' said Dr Joann Elmore, lead author of the study published in the JAMA Network Open journal.

Why do so many companies still struggle to build a smooth-running pipeline from data to insights? They invest in heavily hyped machine-learning algorithms to analyze data and make business predictions. Then, inevitably, they realize that algorithms aren't magic; if they're fed junk data, their insights won't be stellar. So they employ data scientists that spend 90% of their time washing and folding in a data-cleaning laundromat, leaving just 10% of their time to do the job for which they were hired. What is flawed about this process is that companies only get excited about machine learning for end-of-the-line algorithms; they should apply machine learning just as liberally in the early cleansing stages instead of relying on people to grapple with gargantuan data sets, according to Andy Palmer, co-founder and chief executive officer of Tamr Inc., which helps organizations use machine learning to unify their data silos.

Recent advances in supervised machine learning have improved diagnostic accuracy and prediction of treatment outcomes, in some cases surpassing the performance of clinicians.1 In supervised machine learning, a mathematical function is constructed via automated analysis of training data, which consists of input features (such as retinal images) and output labels (such as the grade of macular edema). With large training data sets and minimal human guidance, a computer learns to generalize from the information contained in the training data. The result is a mathematical function, a model, that can be used to map a new record to the corresponding diagnosis, such as an image to grade macular edema. Although machine learning–based models for classification or for predicting a future health state are being developed for diverse clinical applications, evidence is lacking that deployment of these models has improved care and patient outcomes.2 One barrier to demonstrating such improvement is the basis used to assess the performance of a model.

Sudden shrieks of radio waves from deep space keep slamming into radio telescopes on Earth, spattering those instruments' detectors with confusing data. And now, astronomers are using artificial intelligence to pinpoint the source of the shrieks, in the hope of explaining what's sending them to Earth from -- researchers suspect -- billions of light-years across space. Usually, these weird, unexplained signals are detected only after the fact, when astronomers notice out-of-place spikes in their data -- sometimes years after the incident. The signals have complex, mysterious structures, patterns of peaks and valleys in radio waves that play out in just milliseconds. That's not the sort of signal astronomers expect to come from a simple explosion, or any other one of the standard events known to scatter spikes of electromagnetic energy across space. Ever since the first one was uncovered in 2007, using data recorded in 2001, there's been an ongoing effort to pin down their source.

The Trump administration quietly terminated an Obama-era federal committee on automation in transportation earlier this year, the Department of Transportation confirmed to The Verge this week. What's more, the DOT never informed some members that the advisory group didn't exist anymore, including Captain Chesley "Sully" Sullenberger, Zipcar founder Robin Chase, Apple vice president Lisa Jackson, and even the committee's own vice chair, The Verge has learned. The committee's dissolution comes at a critical moment in the development of automated vehicles in the United States. During the two-plus years that it sat dormant, multiple companies have rolled out small commercial fleets of automated vehicles that perform a variety of tasks. Big money is pouring into some of the most visible companies in the space.

Along with sustainable electric transportation, he views autonomy as a core element of Tesla Inc.'s "fundamental goodness." Humans will be freed of the tedium of driving, he told Wall Street last year. Millions of lives will be saved. There is another incentive for Musk to put driverless cars on the road, though. The day he does that, hundreds of millions of dollars' worth of stored-up revenue become eligible for a trip straight to Tesla's perpetually stressed bottom line.

The Food and Drug Administration has allowed medical devices that rely on artificial intelligence algorithms onto the market, but so far, the agency has given the green light only to devices with "locked algorithms" -- those that remain the same as the product is used until they're updated by the manufacturer. Systems with algorithms that evolve and sharpen on their own, however, are already in development. Unlock this article by subscribing to STAT Plus and enjoy your first 30 days free! STAT Plus is STAT's premium subscription service for in-depth biotech, pharma, policy, and life science coverage and analysis. Our award-winning team covers news on Wall Street, policy developments in Washington, early science breakthroughs and clinical trial results, and health care disruption in Silicon Valley and beyond.