We review the task of Sentence Pair Scoring, popular in the literature in various forms - viewed as Answer Sentence Selection, Semantic Text Scoring, Next Utterance Ranking, Recognizing Textual Entailment, Paraphrasing or e.g. a component of Memory Networks. We argue that all such tasks are similar from the model perspective and propose new baselines by comparing the performance of common IR metrics and popular convolutional, recurrent and attention-based neural models across many Sentence Pair Scoring tasks and datasets. We discuss the problem of evaluating randomized models, propose a statistically grounded methodology, and attempt to improve comparisons by releasing new datasets that are much harder than some of the currently used well explored benchmarks. We introduce a unified open source software framework with easily pluggable models and tasks, which enables us to experiment with multi-task reusability of trained sentence model. We set a new state-of-art in performance on the Ubuntu Dialogue dataset.

We present a method based on the orthogonal symmetric non-negative matrix tri-factorization of the normalized Laplacian matrix for community detection in complex networks. While the exact factorization of a given order may not exist and is NP hard to compute, we obtain an approximate factorization by solving an optimization problem. We establish the connection of the factors obtained through the factorization to a non-negative basis of an invariant subspace of the estimated matrix, drawing parallel with the spectral clustering. Using such factorization for clustering in networks is motivated by analyzing a block-diagonal Laplacian matrix with the blocks representing the connected components of a graph. The method is shown to be consistent for community detection in graphs generated from the stochastic block model and the degree corrected stochastic block model. Simulation results and real data analysis show the effectiveness of these methods under a wide variety of situations, including sparse and highly heterogeneous graphs where the usual spectral clustering is known to fail. Our method also performs better than the state of the art in popular benchmark network datasets, e.g., the political web blogs and the karate club data.

Automatic modulation classification (AMC) is an important task for modern communication systems; however, it is a challenging problem when signal features and precise models for generating each modulation may be unknown. We present a new biologically-inspired AMC method without the need for models or manually specified features --- thus removing the requirement for expert prior knowledge. We accomplish this task using regularized stacked sparse denoising autoencoders (SSDAs). Our method selects efficient classification features directly from raw in-phase/quadrature (I/Q) radio signals in an unsupervised manner. These features are then used to construct higher-complexity abstract features which can be used for automatic modulation classification. We demonstrate this process using a dataset generated with a software defined radio, consisting of random input bits encoded in 100-sample segments of various common digital radio modulations. Our results show correct classification rates of > 99% at 7.5 dB signal-to-noise ratio (SNR) and > 92% at 0 dB SNR in a 6-way classification test. Our experiments demonstrate a dramatically new and broadly applicable mechanism for performing AMC and related tasks without the need for expert-defined or modulation-specific signal information.

Dictionary learning has recently received significant attention due to the increased importance of finding sparse representations of signals/data. In dictionary learning, the goal is to construct an overcomplete basis using input signals such that each signal can be described by a small number of atoms (columns) [1]. Although the existing literature has focused on one-dimensional data, many signals in practice are multidimensional and have a tensor structure: examples include 2-dimensional images and 3-dimensional signals produced via magnetic resonance imaging or computed tomography systems. In traditional dictionary learning techniques, multidimensional data are processed after vectorizing of signals. This can result in poor sparse representations as the structure of the data is neglected [2]. In this paper we provide fundamental limits on learning dictionaries for multidimensional data with tensor structure: we call such dictionaries Kronecker-structured (KS). Several algorithms have been proposed to learn KS dictionaries [2]-[7] but there has been little work on the theoretical guarantees of such algorithms. The lower bounds we provide on the minimax risk of learning a KS dictionary give a measure to evaluate the performance of the existing algorithms.

Last week, NVIDIA (NASDAQ:NVDA) released its first-quarter fiscal 2017 report, which included some impressive results. The graphics-processor maker posted Q1 revenue of 1.3 billion, an increase of 13% year over year, and non-GAAP earning per share were up 39% from the year-ago quarter to 0.46. Wall Street analysts had been expecting revenue of around 1.26 billion and EPS of 0.41. NVIDIA co-founder and CEO Jen-Hsun Huang said the revenue and earnings growth were spurred on by all of the key segments of its business. "We are enjoying growth in all of our platforms -- gaming, professional visualization, datacenter and auto," Huang said in a press release.

Truphone co-founder James Tagg recently addressed the critical role artificial intelligence (AI) will play in helping smartphones and other technology platforms evolve. Tagg, who authored'Are The Androids Dreaming Yet,' kicked off his presentation in Mountain View, California, by exploring some of the basic differences between humans and machines. "Human brains are fundamentally'broken' in certain ways, especially when it comes to accurately remembering a specific event a few weeks after it occurred. Yet, we can recall more than 1,000 faces – in less than 370ms (each)," he explained. "We also understand hierarchy quite well, although we have difficulty with recalling names. Plus, humans adhere to strong rules of etiquette around face-to-face communication."

Sure announced today that it has introduced the world's first on-demand insurance app powered by artificial intelligence, marking an important step for the company as it drives rapid adoption of mobile, on-demand insurance products. An industry first, Sure uses data-driven approaches to empower consumers with the ability to personalize insurance needs on the go. In doing so, it eliminates the built-in limitations of an industry driven largely by one-size-fits-all recommendations from brokers. Sure continues to pioneer Episodic Insurance in the insurance market by providing consumers with simple options based on the context and information gathered through his or her mobile device with permission. Sure uses data-intelligence and advanced analytics to identify needs and provide products for easy, on-demand purchase. For the first time, customers can buy the type of insurance they need, for exactly the duration they need it.

SAN FRANCISCO -- When Google kicks off its annual get-together for software developers on Wednesday, expect to hear about the two major trends shaping the future of the technology industry: artificial intelligence and virtual reality. Thousands will descend on Mountain View, Calif., the first time the tech giant is holding the I/O conference in its hometown. Google's conference, sandwiched between Facebook in April and Apple in June, is vying to put on the greatest show on earth for software developers. For weeks, workers have been constructing a Google-themed park to immerse developers in the artificial intelligence-powered future that Google envisions.The outdoor venue, the Shoreline Ampitheatre, is most famous for showcasing the talents of Neil Young, The Who and Metallica. For I/O, the master of ceremonies is Sundar Pichai, Google's newly minted chief executive who will look to dazzle developers with a demo-packed keynote.

Eric, one of the world's first robots, made his public debut in 1928. Before there was Star Wars' C-3PO and the robot who famously warned of "Danger, Will Robinson!" on TV's Lost in Space, there was Eric -- one of the world's first real robots. He was built in 1928, less than a decade after the word "robot" was first used. He wowed audiences in Britain, where he was created, and elsewhere in Europe and the United States. Now, a team from the Science Museum of London is planning to rebuild him, using original archival materials.

Google is set to bring updates to many of its product lines at this year's I/O developer conference Google's annual developer conference I/O (short for "input/output" in computer science lingo) is set to take place this year in Mountain View, California from Wednesday, May 18, through Friday, May 20. And with the arrival of the conference comes the possibility of new products from the search company. While Google is known for releasing products and updates throughout the year, some of its biggest releases are often saved for the I/O conference. Last year's Google I/O 2015 brought fans Android M, Google Cardboard on the iPhone, and Now On Tap -- a smartphone feature that surfaces the right information at the right time. We can't know for sure what the company (which is now technically a subsidiary of the larger conglomerate known as Alphabet) will announce, but some hints have dropped.