This work proposes a procedure for designing algorithms for specific adaptive data collection tasks like active learning and pure-exploration multi-armed bandits. Unlike the design of traditional adaptive algorithms that rely on concentration of measure and careful analysis to justify the correctness and sample complexity of the procedure, our adaptive algorithm is learned via adversarial training over equivalence classes of problems derived from information theoretic lower bounds. In particular, a single adaptive learning algorithm is learned that competes with the best adaptive algorithm learned for each equivalence class. Our procedure takes as input just the available queries, set of hypotheses, loss function, and total query budget. This is in contrast to existing meta-learning work that learns an adaptive algorithm relative to an explicit, user-defined subset or prior distribution over problems which can be challenging to define and be mismatched to the instance encountered at test time. This work is particularly focused on the regime when the total query budget is very small, such as a few dozen, which is much smaller than those budgets typically considered by theoretically derived algorithms. We perform synthetic experiments to justify the stability and effectiveness of the training procedure, and then evaluate the method on tasks derived from real data including a noisy 20 Questions game and a joke recommendation task.

We've heard it for years: 5G is coming. And yet, while high-speed 5G internet has indeed slowly been rolling out in a smattering of countries across the globe, many barriers remain that have prevented widespread adoption. One issue is that we can't get faster internet speeds without more efficient ways of delivering wireless signals. The general trend has been to simply add antennas to either the transmitter (i.e., Wi-Fi access points and cell towers) or the receiver (such as a phone or laptop). But that's grown difficult to do as companies increasingly produce smaller and smaller devices, including a new wave of "internet of things" systems.

Before Insurance Australia Group (IAG) can begin selling more than insurance products to consumers, the company realised it needed to shift what is currently a very analogue business into something that is more digitally orientated. "We really want to change the mindset, get some records and customer value, and build new businesses as well, including beyond insurance," IAG Digital Architecture director Ian Jamieson explained at New Relic Future Stack 2019 last week. "An insurance company not selling insurance is quite disruptive because a lot of the core systems is under insurance, so if we want to sell a solution that provides emergency assistance to your home that is not an insurance product, how would you bill someone for that so that it's not under insurance … there are a range of things we need to transform and add new capabilities to." Jamieson said some areas that IAG is looking to expand its business into include motor and home repair services, spinning up brand new businesses such as in mobility services, and through acquisitions of startups, such as its most recent purchase of Carbar, a subscription-based car ownership platform. To make sure these business plans become a reality, Jamieson said the company has moved away from taking traditional waterfall approaches to projects and using a cross-functional method.

Existing ordinal embedding methods usually follow a two-stage routine: outlier detection is first employed to pick out the inconsistent comparisons; then an embedding is learned from the clean data. However, learning in a multi-stage manner is well-known to suffer from sub-optimal solutions. In this paper, we propose a unified framework to jointly identify the contaminated comparisons and derive reliable embeddings. The merits of our method are three-fold: (1) By virtue of the proposed unified framework, the sub-optimality of traditional methods is largely alleviated; (2) The proposed method is aware of global inconsistency by minimizing a corresponding cost, while traditional methods only involve local inconsistency; (3) Instead of considering the nuclear norm heuristics, we adopt an exact solution for rank equality constraint. Our studies are supported by experiments with both simulated examples and real-world data. The proposed framework provides us a promising tool for robust ordinal embedding from the contaminated comparisons.

This paper considers a multi-armed bandit game where the number of arms is much larger than the maximum budget and is effectively infinite. We characterize necessary and sufficient conditions on the total budget for an algorithm to return an {\epsilon}-good arm with probability at least 1 - {\delta}. In such situations, the sample complexity depends on {\epsilon}, {\delta} and the so-called reservoir distribution {\nu} from which the means of the arms are drawn iid. While a substantial literature has developed around analyzing specific cases of {\nu} such as the beta distribution, our analysis makes no assumption about the form of {\nu}. Our algorithm is based on successive halving with the surprising exception that arms start to be discarded after just a single pull, requiring an analysis that goes beyond concentration alone. The provable correctness of this algorithm also provides an explanation for the empirical observation that the most aggressive bracket of the Hyperband algorithm of Li et al. (2017) for hyperparameter tuning is almost always best.

Airborne data collecting platforms like the RQ-4 Global Hawk have a problem: the usefulness of the data they collect is limited by how fast and how well it can be analyzed. US military intelligence gathers a lot of data, but in order to make the data useful for a decision making process, the Air Force needs a "sensing grid that fuses together data," C4ISRNET reported Wednesday. AI will help the force interpret that fused data. The AI will harvest information from airborne systems in development such as Gremlin drones, which the US military portrays as a swarm of small drones that take off from an aircraft mid-flight and are recovered by the same aircraft. "How do I get the data so I can fuse it, look at it and then ask the right questions from the data to reveal what trends are out there?" Lt. Gen. VeraLinn Jamieson said in a July 31 interview with the news outlet.

China's massive investment in artificial intelligence technologies may soon leave the U.S. at a major disadvantage, a top Air Force general said Thursday. "Speed is of the essence in the digital age," said Lt. Gen. VeraLinn "Dash" Jamieson, deputy chief of staff for intelligence, surveillance and reconnaissance on the Air Staff at the Pentagon. She painted a grim picture: While "great instigator" Russia has the desire to do ambitious experiments with A.I., China already has the means. For example, China is building several digital artificial intelligence cities in a military-civilian partnership to understand how A.I. will be propagated as it strives to become the global leader in technology. The cities track human movement through artificial facial recognition software, watching citizens' every move as they go about their day.

China's progress towards its goal of becoming the world's leader in AI by the year 2025 remains unchecked. While its efforts still lag behind the US, thanks to the likes of Google and Microsoft, there's an alarming amount of research indicating the gap is shrinking. It's only been a year since TNW reported China's announcement it was shifting its national strategy to claim the artificial intelligence crown. In that time China has advanced its agenda to a startling degree, at least according to the experts. Air Force General VeraLinn "Dash" Jamieson, deputy chief of staff for intelligence, surveillance and reconnaissance on the Air Staff at the Pentagon, spoke last Thursday at a military event. She told attendees she was concerned that China is taking AI more seriously than other nations.

The Air Force is hoping advances in artificial intelligence mean the abundance of data from its aircraft, weapons and satellites will be easier to access and analyze. "There's certainly a data aspect piece" of artificial intelligence, Air Force Vice Chief of Staff Gen. Stephen "Seve" Wilson said Monday. "I think we all would agree, maybe, that data is the'renewable oil' of the 21st century," Wilson said during the Future of War conference hosted by New America and Arizona State University. "With that data, we have to be able to have the right algorithms that connect the data, and understand the data, to a network. I think the cloud's a part of this, and I think'compute on the edge' is a part of this," he said, referencing a streamlined approach to avoid stovepiping information flow needed at the speed of war.

Performance of machine learning algorithms depends critically on identifying a good set of hyperparameters. While current methods offer efficiencies by adaptively choosing new configurations to train, an alternative strategy is to adaptively allocate resources across the selected configurations. We formulate hyperparameter optimization as a pure-exploration non-stochastic infinitely many armed bandit problem where a predefined resource like iterations, data samples, or features is allocated to randomly sampled configurations. We introduce Hyperband for this framework and analyze its theoretical properties, providing several desirable guarantees. Furthermore, we compare Hyperband with state-of-the-art methods on a suite of hyperparameter optimization problems. We observe that Hyperband provides five times to thirty times speedup over state-of-the-art Bayesian optimization algorithms on a variety of deep-learning and kernel-based learning problems.