A system of nested dichotomies is a method of decomposing a multi-class problem into a collection of binary problems. Such a system recursively applies binary splits to divide the set of classes into two subsets, and trains a binary classifier for each split. Although ensembles of nested dichotomies with random structure have been shown to perform well in practice, using a more sophisticated class subset selection method can be used to improve classification accuracy. We investigate an approach to this problem called random-pair selection, and evaluate its effectiveness compared to other published methods of subset selection. We show that our method outperforms other methods in many cases when forming ensembles of nested dichotomies, and is at least on par in all other cases.

Linear dimensionality reduction techniques are powerful tools for image analysis as they allow the identification of important features in a data set. In particular, nonnegative matrix factorization (NMF) has become very popular as it is able to extract sparse, localized and easily interpretable features by imposing an additive combination of nonnegative basis elements. Nonnegative matrix underapproximation (NMU) is a closely related technique that has the advantage to identify features sequentially. In this paper, we propose a variant of NMU that is particularly well suited for image analysis as it incorporates the spatial information, that is, it takes into account the fact that neighboring pixels are more likely to be contained in the same features, and favors the extraction of localized features by looking for sparse basis elements. We show that our new approach competes favorably with comparable state-of-the-art techniques on synthetic, facial and hyperspectral image data sets.

In this paper, we propose a clustering-based iterative algorithm to solve certain optimization problems in machine learning when data size is large and thus it becomes impractical to use out-of-the-box algorithms. We rely on the principle of data aggregation and then subsequent disaggregations. While it is standard practice to aggregate the data and then calibrate the machine learning algorithm on aggregated data, we embed this into an iterative framework where initial aggregations are gradually disaggregated to the extent that even an optimal solution is obtainable. Early studies in data aggregation consider transportation problems [1, 10], where either demand or supply nodes are aggregated. Zipkin [31] studied data aggregation for linear programming (LP) and derived error bounds of the approximate solution.

Tesla Motors shipped fewer autos to customers in the last three months, making it unlikely to meet prior expectations for delivering 80,000 to 90,000 vehicles this year. The Palo Alto maker of electric autos said Sunday that it delivered 14,370 vehicles in the April to June quarter, a decline of 450 vehicles from the first quarter that Tesla attributed to an "extreme production ramp up" and a number of custom-ordered vehicles still being shipped. Tesla said it anticipates delivering 50,000 vehicles in the second half of the year. Although that second-half target would match its vehicle deliveries for all of 2015, it would still be just shy of the guidance provided by the company in April. The revised expectations arrive at a delicate moment for Tesla, which has excited drivers and investors alike by the promise of gasoline-free autos and drummed up significant hype over its upcoming Model 3, pitched as an electric car for the masses.

This is the text version of remarks I gave on June 26, 2016, at a panel on the Moral Economy of Tech at the SASE conference in Berkeley. The other panel participants were Kieran Healy, Stuart Russell and AnnaLee Saxenian. We were each asked to speak for ten minutes, to an audience of social scientists. I am only a small minnow in the technology ocean, but since it is my natural habitat, I want to make an effort to describe it to you. As computer programmers, our formative intellectual experience is working with deterministic systems that have been designed by other human beings. These can be very complex, but the complexity is not the kind we find in the natural world.

Pulling these threads together, recently a gaming AI defeated a US Air Force expert combat pilot in simulated combat.

There is lots of talk about machine learning at the moment, says Alexander Graubner-Müller. AI can write songs, compose novels and even beat the world champion at Go – and machine learning can even help with financial services. Consumer credit is one of these areas, says Graubner-Müller, who cofounded Kreditech in 2012. There are two classes of consumer credit, he says – people with no access to credit and people with access to credit. One is middle class, well-employed and has a strong credit history.

LG Electronics and Incheon International Airport Corporation have joined forces to upgrade the nation's largest international airport to what is billed as a smarter and more tech-savvy one. The two parties signed a memorandum of understanding Friday which says LG Electronics will deploy automated robots and Internet of Things technology to facilities at the airport. Among the new services are robots that will stroll around the airport giving directions to airport newcomers. The guide service will be provided in Korean, English, Japanese and Chinese. Some of the robots will clean the floors, organize luggage and check security.

The history of artificial intelligence (AI) dates back into antiquity – intelligent robots appear in the myths of many ancient societies, including Greek, Arabic, Egyptian and Chinese. Today, the field of artificial intelligence is more vibrant than ever and some believe that we're on the threshold of discoveries that could change human society irreversibly, for better or worse. Humans tend to think in straight lines, but every aspect of technological progress is actually accelerating – including AI. Futurist Ray Kurzweil calls this the "Law of Accelerating Returns", and presents evidence that an amount of progress equal to the entire 20th century's gains was attained between 2000 and 2014. He also argues that the same amount will happen again before 2021. Understanding the exponential nature of progress and ignoring the inner tendency to think things will keep improving at the same rate is key to getting to grips with how fast we'll make scientific advances in the future.

NEW DELHI: Panasonic India is scouting for companies to acquire in 6-9 months to develop artificial intelligence (AI) and Machine learning technologies, which it wants to integrate with its future smartphones to differentiate from rival vendors in the crowded yet fast growing market. The handset vendor has already set aside an initial corpus of 10 million for the development of this technology through a merger and acquisition or a joint venture. "The budget is in tune of 10 million to start with, and as we see progress on this front and things go in right direction, then there will be no constraint on the budget part. We can spend as high as possible. Some part of this budget has been generated from the India business, while some portion has been allocated from Japan," Pankaj Rana, head of mobility division, India, South Asia, Middle East and Africa at Panasonic, told ET. "Our team would be traveling to Silicon Valley soon. We will have new products ready with AI in 9-12 months. In the last three months, we have finalized what we will do and budgets have already been allocated from Panasonic Japan and Panasonic India. Now we have to find partner and start executive on timeline, while understanding the market," he said.