One of the ways to train deep neural networks effectively is to use residual connections. Residual connections can be classified as being either identity connections or bridge-connections with a reshaping convolution. Empirical observations on CIFAR-10 and CIFAR-100 datasets using a baseline Resnet model, with bridge-connections removed, have shown a significant reduction in accuracy. This reduction is due to lack of contribution, in the form of feature maps, by the bridge-connections. Hence bridge-connections are vital for Resnet. However, all feature maps in the bridge-connections are considered to be equally important. In this work, an upgraded architecture "Res-SE-Net" is proposed to further strengthen the contribution from the bridge-connections by quantifying the importance of each feature map and weighting them accordingly using Squeeze-and-Excitation (SE) block. It is demonstrated that Res-SE-Net generalizes much better than Resnet and SE-Resnet on the benchmark CIFAR-10 and CIFAR-100 datasets.

Deep learning has achieved remarkable successes in solving challenging reinforcement learning (RL) problems when dense reward function is provided. However, in sparse reward environment it still often suffers from the need to carefully shape reward function to guide policy optimization. This limits the applicability of RL in the real world since both reinforcement learning and domain-specific knowledge are required. It is therefore of great practical importance to develop algorithms which can learn from a binary signal indicating successful task completion or other unshaped, sparse reward signals. We propose a novel method called competitive experience replay, which efficiently supplements a sparse reward by placing learning in the context of an exploration competition between a pair of agents. Our method complements the recently proposed hindsight experience replay (HER) by inducing an automatic exploratory curriculum. We evaluate our approach on the tasks of reaching various goal locations in an ant maze and manipulating objects with a robotic arm. Each task provides only binary rewards indicating whether or not the goal is achieved. Our method asymmetrically augments these sparse rewards for a pair of agents each learning the same task, creating a competitive game designed to drive exploration. Extensive experiments demonstrate that this method leads to faster converge and improved task performance.

Abstract--In order to maximize detection precision rate as well as the recall rate, this paper proposes an in-vehicle multisource fusionscheme in Keyword Spotting (KWS) System for vehicle applications. Vehicle information, as a new source for the original system, is collected by an in-vehicle data acquisition platform while the user is driving. A Deep Neural Network (DNN) is trained to extract acoustic features and make a speech classification. Based on the posterior probabilities obtained from DNN, the vehicle information including the speed and direction of vehicle is applied to choose the suitable parameter from a pair of sensitivity values for the KWS system. The experimental results show that the KWS system with the proposed multi-source fusion scheme can achieve better performances in term of precision rate, recall rate, and mean square error compared to the system without it. I. INTRODUCTION Keyword Spotting (KWS) System, also known as wakeword detection,refers to the task of detecting specified keyword from a continuous stream of audio provided by the users [1]. Keyword Spotting has been an active research area in speech recognition for decades, and widely used in numerous applications.

AI research quickly accelerated, with Kunihiko Fukushima developing the first true, multilayered neural network in 1975. The original goal of the neural network approach was to create a computational system that could solve problems like a human brain. However, over time, researchers shifted their focus to using neural networks to match specific tasks, leading to deviations from a strictly biological approach. Since then, neural networks have supported diverse tasks, including computer vision, speech recognition, machine translation, social network filtering, playing board and video games, and medical diagnosis. As structured and unstructured data sizes increased to big data levels, people developed deep learning systems, which are essentially neural networks with many layers.

IBM wants to develop next-generation artificial intelligence chips, and it's building a new AI research center and partnering with academia and other tech companies to do it. At the recently announced future AI Hardware Center at SUNY Polytechnical Institute in Albany, New York, IBM researchers will collaborate with academic researchers and tech partners to develop, prototype, and test new AI chips and systems. Initial partners include Samsung, Mellanox Technologies, Synopsis, Applied Materials, and Tokyo Electron. Related: Intel Steps Up Its Challenge to Nvidia's AI Chip Dominance, with Facebook's Help The IBM Research division, which has designed several prototypes of its Digital AI cores and Analog AI cores in recent years, will continue to develop these chips at the center, Jeff Burns, IBM Research's director of AI Compute and director of the future AI Hardware Center, said. These new processors are expected to result in a 1,000-times improvement in AI compute performance efficiency over the next 10 years.

The Clippers and Second Spectrum will use AWS machine learning and data analytics services to advance game analyses and drive new experiences for Clippers CourtVision, which launched to great acclaim at the start of the 2018-19 basketball season and has been billed by experts as the future of sports viewing. In addition, Clippers CourtVision will test Amazon SageMaker to build, train, and deploy machine learning-driven stats which will appear on live broadcasts and on-demand NBA game videos. Second Spectrum uses cameras in all 29 NBA arenas to collect 3D spatial data including ball and player locations and movements, which is stored and analyzed on AWS in real time. With help from AWS's broad range of services, Second Spectrum uses that data to generate augmented graphical overlays on Clippers broadcasts in real time, offering users an array of content options and Clippers CourtVision Modes with features ranging from live layouts of basketball plays, to the frame-by-frame probability of a shot going in, to a suite of graphics that animate based on conditions both simple and complex, giving fans a deeper understanding of and interaction with the game as the action unfurls on the court. Clippers CourtVision uses AWS Elemental Media Services to deliver the live game-watching experience.

At Callaway, the high-end golf-equipment stalwart, the process of making clubs has always been quite labor-intensive--from grinding and polishing clubheads to crafting wood-and-steel-shafted irons and wedges. The company has also long combined such artisanal handwork with technological innovation, even partnering with aerospace titan Boeing recently to codesign several aerodynamic clubs. So when the company set out about four years ago to make its latest club line, called Epic Flash, it took the next evolutionary technological step, turning to artificial intelligence and machine learning for help. A typical club-design process might involve five to seven physical prototypes; for Epic Flash, Callaway created 15,000 virtual ones. From those, an algorithm determined the best design, selecting for peak performance--i.e., ball speed--while also conforming to the rules set forth by the U.S. Golf Association.

A popular dating app has become the latest victim of a major data breach after hackers exposed the details of 6 million of its users. Hacked information of Coffee Meets Bagel users appeared in a huge cache of data that appeared on a popular dark web marketplace earlier this week. The previously undisclosed breach has since been acknowledged by the dating app. Coffee Meets Bagel revealed details about the hack in an email to its users on Valentine's Day, explaining that members' names and email addresses had been exposed. "We recently discovered that some data from your Coffee Meets Bagel account may have been acquired by an unauthorised party," the email stated.

Dubai International Airport is the latest to halt flights over a drone scare following similar incidents at London's Gatwick and Heathrow. The world's third-busiest airport temporarily stopped operations for just under 30 minutes due to "unauthorized drone activity," according to a tweet from the Dubai Media Office. Incoming flights were permitted to land during the disruption, reports The New York Times, which occurred between 10.15AM and 10.45AM local time. Operations are now reportedly back to normal. "Dubai Airports has worked closely with the appropriate authorities to ensure that the safety of airport operations is maintained at all times and to minimize any inconvenience to our customers," the airport said.

Amgen's drug discovery group is a few blocks beyond that. Until recently, Barzilay, one of the world's leading researchers in artificial intelligence, hadn't given much thought to these nearby buildings full of chemists and biologists. But as AI and machine learning began to perform ever more impressive feats in image recognition and language comprehension, she began to wonder: could it also transform the task of finding new drugs? The problem is that human researchers can explore only a tiny slice of what is possible. It's estimated that there are as many as 1060 potentially drug-like molecules--more than the number of atoms in the solar system. But traversing seemingly unlimited possibilities is what machine learning is good at. Trained on large databases of existing molecules and their properties, the programs can explore all possible related molecules.