We introduce the speculate-correct method to derive error bounds for local classifiers. Using it, we show that k-nearest neighbor classifiers, in spite of their famously (fractured decision boundaries, have exponential error bounds (k) with O lnn)/n error bound range for n in-sample examples. Keywords: nearest neighbors, statistical learning, supervised learning, error bounds, generalization 2000 MSC: 62G99, 2000 MSC: 68Q32, 2000 MSC: 62M99 1. Introduction Local classifiers use only a small subset of their examples to classify each input. The best-known local classifier is the nearest neighbor classifier. To classify an example, a k-nearest neighbor (k-nn) classifier uses a majority vote over the k in-sample examples closest to the example. We assume k is odd, and we assume binary classification.

We propose a method for hand pose estimation based on a deep regressor trained on two different kinds of input. Raw depth data is fused with an intermediate representation in the form of a segmentation of the hand into parts. This intermediate representation contains important topological information and provides useful cues for reasoning about joint locations. The mapping from raw depth to segmentation maps is learned in a semi/weakly-supervised way from two different datasets: (i) a synthetic dataset created through a rendering pipeline including densely labeled ground truth (pixelwise segmentations); and (ii) a dataset with real images for which ground truth joint positions are available, but not dense segmentations. Loss for training on real images is generated from a patch-wise restoration process, which aligns tentative segmentation maps with a large dictionary of synthetic poses. The underlying premise is that the domain shift between synthetic and real data is smaller in the intermediate representation, where labels carry geometric and topological meaning, than in the raw input domain. Experiments on the NYU dataset show that the proposed training method decreases error on joints over direct regression of joints from depth data by 15.7%.

DOI: 10.18713/JIMIS-010917-3-1 Submitted: 12/2/2016 - Published: 6/2/2017 Volume: 3 - Year: 2017 Issue: Digital Contextualization Editors: Frédéric Lebaron, Brigitte Le Roux, Fionn Murtagh, Evelyn Ruppert The relevance and importance of contextualizing data analytics is described. Qualitative characteristics might form the context of quantitative analysis. Topics that are at issue include: contrast, baselining, secondary data sources, supplementary data sources; dynamic and heterogeneous data. In geometric data analysis, especially with the Correspondence Analysis platform, various case studies are both experimented with, and are reviewed. In such aspects as paradigms followed, and technical implementation, implicitly and explicitly, an important point made is the major relevance of such work for both burgeoning analytical needs and for new analytical areas including Big Data analytics, and so on. For the general reader, it is aimed to display and describe, first of all, the analytical outcomes that are subject to analysis here, and then proceed to detail the more quantitative outcomes that fully support the analytics carried out.

IMAGE: An international team of scientists presents a thorough review on quantum machine learning, its current status and future prospects. The reports contrasts machine learning using classical and quantum resources, identifying... view more Language acquisition in young children is apparently connected with their ability to detect patterns. In their learning process, they search for patterns in the data set that help them identify and optimize grammar structures in order to properly acquire the language. Likewise, online translators use algorithms through machine learning techniques to optimize their translation engines to produce well-rounded and understandable outcomes. Even though many translations did not make much sense at all at the beginning, in these past years we have been able to see major improvements thanks to machine learning.

The era of artificial intelligence is upon us, though there's plenty of debate over how AI should be defined much less whether we should start worrying about an apocalyptic robot uprising. The latter issue recently ignited a highly publicized dispute between Elon Musk and Mark Zuckerberg, who argued that it was irresponsible to "try to drum up these doomsday scenarios". In the near-term however, it seems more than likely that AI will be weaponized by hackers in criminal organizations and governments to enhance now-familiar forms of cyberattacks like identity theft and DDoS attacks. A recent survey has found that a majority of cybersecurity professionals believe that artificial intelligence will be used to power cyberattacks in the coming year. Cybersecurity firm Cylance conducted the survey at this year's Black Hat USA conference and found that 62 percent of respondents believe that "there is high possibility that AI could be used by hackers for offensive purposes."

Of the many, many (many many many) challenges that are inherent to urban drone delivery, safety is one of the most important. Nobody has a reliable, cost-effective solution for this, although we've seen some unreliable ones (dangling packages on strings) and cumbersome ones (dedicated, protected landing pads), so we've been missing an elegant way of protecting end users from robots that fly with spinning blades of death. EPFL in Switzerland has had a solution for this for years--drones surrounded by protective cages that allow them to bounce off of obstacles. As far as the drones are concerned, humans are obstacles as well, so a protective cage does pretty well at protecting them from us (and vice versa). The annoying thing about these cages has always been that they're all kinds of bulky, especially if they're protecting a quadrotor beefy enough to be useful.

Miniature robots with arms and legs made of DNA can sort and deliver molecular cargo, a new study finds. Such DNA robots could be used to shuffle nanoparticles around on circuits, assemble therapeutic compounds, separate molecular components in trash for recycling, or deliver medicines where they need to go in the body, researchers from the California Institute of Technology in Pasadena say. "Just like electromechanical robots have been sent to places that are perhaps too far for humans to go to--for example, on another planet--if we truly master the ways of engineering molecular machines, we would be able to build molecular robots and send them to places that are perhaps too small for humans to go to--for example, inside the bloodstream," says study senior author Lulu Qian, an assistant professor of bioengineering at Caltech. The new robots are made from three basic modules that are each brief snippets of DNA. One module is a "leg" with two "feet" for walking; another is an "arm" with a "hand" for grabbing onto cargo; and the last can recognize specific delivery points and make the hand release its cargo at those spots.

One night in late July 2014, a journalist from the Chinese newspaper Southern Weekly interviewed a 17-year-old Chinese girl named Xiaoice (pronounced Shao-ice). The journalist, Liu Jun, conducted the interview online, through the popular social networking platform Weibo. LJ: So many people make fun of you and insult you, why don't you get mad? Xiaoice: You should ask my father. LJ: What if your father leaves you one day unattended?

Facebook announced today that it is using artificial intelligence to make sure that 360-degree photos uploaded to the social network look their best when other people view them. The company laid out a system at its @Scale conference today that uses deep neural networks to try to correct for common orientation errors with the photos that are uploaded. If someone taking a 360 degree photo doesn't hold the camera perfectly in line with the horizon, the resulting image can be tilted, which makes it harder to read and breaks the sense of immersion if the image is being viewed in virtual reality. Facebook's system takes in a photo and outputs a pair of values for the tilt and roll correction needed to bring the horizon of the photo in line. That way, it doesn't feel like users are viewing a crooked image when they look around a scene.

One common task in NLP (Natural Language Processing) is tokenization. "Tokens" are usually individual words (at least in languages like English) and "tokenization" is taking a text or set of text and breaking it up into its individual words. These tokens are then used as the input for other types of analysis or tasks, like parsing (automatically tagging the syntactic relationship between words). In this tutorial you'll learn how to: For this tutorial we'll be using a corpus of transcribed speech from bilingual children speaking in English. You can find more information on this dataset and download it here.