Self-driving cars will rarely have to deal with a pack of drivers who think they are in a "Fast and Furious" movie, but training them to do so might just be what it takes to reach true autonomy. That's why Ascent Robotics Inc. is building a virtual simulation that it believes will help create self-driving automobiles capable of handling any scenario, however unlikely. The Tokyo-based startup is raising ¥1.1 billion ($10 million) in its first funding round, led by SBI Investment Co. The total distance traveled by driverless vehicles on public roads has long been considered the main metric of progress in the industry. By that measure alone, the 8 million km (5 million miles) logged by Alphabet Inc.'s Waymo would appear to be an insurmountable lead.

Facebook has a problem it just can't kick: People keep exploiting it in ways that could sway elections, and in the worst cases even undermine democracy. News reports that Facebook let the Trump-affiliated data mining firm Cambridge Analytica abscond with data from tens of millions of users mark the third time in roughly a year the company appears to have been outfoxed by crafty outsiders in this way. Before the Cambridge imbroglio, there were Russian agents running election-related propaganda campaigns through targeted ads and fake political events. And before the Russians took centre stage, there were purveyors of fake news who spread false stories to rile up hyper-partisan audiences and profit from the resulting ad revenue. In the previous cases, Facebook initially downplayed the risks posed by these activities.

EDIT* This guide was written for fastai version 1, which at the current date and time (Jan 2018) is in the midst of transitioning to the a newer version, dubbed fastai v2. An updated guide will be coming soon. As a deep learning enthusiast in Malaysia, one of the biggest issues I have is securing a cheap GPU option to run my models on. If you're like me and come from a country where paying $80-$100 every month for AWS GPUs is too expensive, I'll show you how I set up a my GPU on GCP without incurring any cost at all. In this article I'll walk you through setting up a google cloud computing instance with a 500gb SSD, a 3.75gb ram Broadwell CPU and a Nvidia Tesla K80 GPU.

It's amazing to see what technology can do these days! Satellites provide daily images of almost every location on earth, and computers can be trained to process massive amounts of data generated from them to produce insightful analysis/information. This is just one of the demonstrations of artificial intelligence (AI). AI can go beyond just reading images captured from space, it can help improve lives overall. For urban governance, machine learning and AI are increasingly used to provide near real-time analysis of how cities change in practice – for example, through the conversion of green areas into built-up structures.

NTT DoCoMo has launched an artificial intelligence (AI) engine that analyzes shelf allocation in stores and warehouses using photos taken with smartphones and other common devices. The image recognition engine employs DoCoMo's AI technology and constitutes part of NTT Group's corevo AI technology. Object-detection technology detects individual items in an image with over 98% accuracy and object-recognition technology identifies specific products with over 95% accuracy by matching them with images stored in a database. Currently, shelf-analysis technology requires products to be placed in the front row and facing forward to ensure high-precision recognition. DoCoMo says its new engine can recognize products on shelves without special arrangement, even when they are packed tightly together.

In a ceremony announcing the launch of the Center for Computational Criminology in Beersheva, Israel Police Commissioner Roni Alsheikh and Ben-Gurion University President Rivka Carmi revealed cooperative plans to fight cyber crime with cutting-edge technology. The center, located at BGU's Advanced Technologies Park, will develop advanced cyber, big data and artificial intelligence tools to fight cyber crime, which has risen exponentially in recent years as criminals and even rogue governments have capitalized on the anonymity of cyberspace to cloak their activities while reaping sizeable profits. BGU researchers will work together with the Israel Police's cyber investigators to develop new artificial intelligence and machine learning tools for law enforcement. "The last, most significant scientific breakthrough to change law enforcement was DNA testing," said the head of the new center, Prof. Lior Rokach, chairman of BGU's Department of Software and Information Systems Engineering, and a leading expert on artificial intelligence. "Today, we are on the threshold of the next big breakthrough: analyzing big data to discover hidden patterns to predict and prevent crime. The AI revolution of the past few years will prove to be even more significant than DNA testing for law enforcement, providing them with unprecedented investigative tools and new sources of evidence."

A new Bayesian modeling method is proposed by combining the maximization of the marginal likelihood with a momentum-space renormalization group transformation for Gaussian graphical models. Moreover, we present a scheme for computint the statistical averages of hyperparameters and mean square errors in our proposed method based on a momentumspace renormalization transformation.

In this paper we examine data fusion methods for multi-view data classification. We present a decision concept which explicitly takes into account the input multi-view structure, where for each case there is a different subset of relevant views. This data fusion concept, which we dub Mixture of Views, is implemented by a special purpose neural network architecture. It is demonstrated on the task of classifying breast microcalcifications as benign or malignant based on CC and MLO mammography views. The single view decisions are combined by a data-driven decision, according to the relevance of each view in a given case, into a global decision. The method is evaluated on a large multi-view dataset extracted from the standardized digital database for screening mammography (DDSM). The experimental results show that our method outperforms previously suggested fusion methods.

This paper presents a statistical method of single-channel speech enhancement that uses a variational autoencoder (VAE) as a prior distribution on clean speech. A standard approach to speech enhancement is to train a deep neural network (DNN) to take noisy speech as input and output clean speech. Although this supervised approach requires a very large amount of pair data for training, it is not robust against unknown environments. Another approach is to use non-negative matrix factorization (NMF) based on basis spectra trained on clean speech in advance and those adapted to noise on the fly. This semi-supervised approach, however, causes considerable signal distortion in enhanced speech due to the unrealistic assumption that speech spectrograms are linear combinations of the basis spectra. Replacing the poor linear generative model of clean speech in NMF with a VAE---a powerful nonlinear deep generative model---trained on clean speech, we formulate a unified probabilistic generative model of noisy speech. Given noisy speech as observed data, we can sample clean speech from its posterior distribution. The proposed method outperformed the conventional DNN-based method in unseen noisy environments.

Exponential families and mixture families are parametric probability models that can be geometrically studied as smooth statistical manifolds with respect to any statistical divergence like the Kullback-Leibler (KL) divergence or the Hellinger divergence. When equipping a statistical manifold with the KL divergence, the induced manifold structure is dually flat, and the KL divergence between distributions amounts to an equivalent Bregman divergence on their corresponding parameters. In practice, the corresponding Bregman generators of mixture/exponential families require to perform definite integral calculus that can either be too time-consuming (for exponentially large discrete support case) or even do not admit closed-form formula (for continuous support case). In these cases, the dually flat construction remains theoretical and cannot be used by information-geometric algorithms. To bypass this problem, we consider performing stochastic Monte Carlo (MC) estimation of those integral-based mixture/exponential family Bregman generators. We show that, under natural assumptions, these MC generators are almost surely Bregman generators. We define a series of dually flat information geometries, termed Monte Carlo Information Geometries, that increasingly-finely approximate the untractable geometry. The advantage of this MCIG is that it allows a practical use of the Bregman algorithmic toolbox on a wide range of probability distribution families. We demonstrate our approach with a clustering task on a mixture family manifold.