The graph Laplacian plays key roles in information processing of relational data, and has analogies with the Laplacian in differential geometry. In this paper, we generalize the analogy between graph Laplacian and differential geometry to the hypergraph setting, and propose a novel hypergraph $p$-Laplacian. Unlike the existing two-node graph Laplacians, this generalization makes it possible to analyze hypergraphs, where the edges are allowed to connect any number of nodes. Moreover, we propose a semi-supervised learning method based on the proposed hypergraph $p$-Laplacian, and formalize them as the analogue to the Dirichlet problem, which often appears in physics. We further explore theoretical connections to normalized hypergraph cut on a hypergraph, and propose normalized cut corresponding to hypergraph $p$-Laplacian. The proposed $p$-Laplacian is shown to outperform standard hypergraph Laplacians in the experiment on a hypergraph semi-supervised learning and normalized cut setting.

Convolutional Neural Networks (CNN) and the locally connected layer are limited in capturing the importance and relations of different local receptive fields, which are often crucial for tasks such as face verification, visual question answering, and word sequence prediction. To tackle the issue, we propose a novel locally smoothed neural network (LSNN) in this paper. The main idea is to represent the weight matrix of the locally connected layer as the product of the kernel and the smoother, where the kernel is shared over different local receptive fields, and the smoother is for determining the importance and relations of different local receptive fields. Specifically, a multi-variate Gaussian function is utilized to generate the smoother, for modeling the location relations among different local receptive fields. Furthermore, the content information can also be leveraged by setting the mean and precision of the Gaussian function according to the content. Experiments on some variant of MNIST clearly show our advantages over CNN and locally connected layer.

Networked data, in which every training example involves two objects and may share some common objects with others, is used in many machine learning tasks such as learning to rank and link prediction. A challenge of learning from networked examples is that target values are not known for some pairs of objects. In this case, neither the classical i.i.d.\ assumption nor techniques based on complete U-statistics can be used. Most existing theoretical results of this problem only deal with the classical empirical risk minimization (ERM) principle that always weights every example equally, but this strategy leads to unsatisfactory bounds. We consider general weighted ERM and show new universal risk bounds for this problem. These new bounds naturally define an optimization problem which leads to appropriate weights for networked examples. Though this optimization problem is not convex in general, we devise a new fully polynomial-time approximation scheme (FPTAS) to solve it.

An artificial intelligence enabled robot has passed the written test of China's national medical licensing examination for the first time, marking another milestone in the quest for AI technology to match or surpass human intelligence. Named Xiaoyi, the robot developed by Tsinghua University and Chinese information technology firm iFlytek, achieved a score of 456, 96 points higher than the required mark of 360 points, according to a company announcement. To pass the text, Xiaoyi was required to memorize and understand the contents of one million medical images, 53 medical books, two million medical records, and 400,000 pieces of medical literature and medical reports, a task which normally takes five years of study by a medical student. The robot reportedly failed an earlier attempt to pass the test. "Xiaoyi's successful pass in the written exam represent a significant development in the field of cognitive intelligence," said iFlytek in the company announcement.

Intelligence agencies have a limited number of trained human analysts looking for undeclared nuclear facilities, or secret military sites, hidden among terabytes of satellite images. But the same sort of deep learning artificial intelligence that enables Google and Facebook to automatically filter images of human faces and cats could also prove invaluable in the world of spy versus spy. An early example: US researchers have trained deep learning algorithms to identify Chinese surface-to-air missile sites--hundreds of times faster than their human counterparts. The deep learning algorithms proved capable of helping people with no prior imagery analysis experience find surface-to-air missile sites scattered across nearly 90,000 square kilometers of southeastern China. Such AI based on neural networks--layers of artificial neuron capable of filtering and learning from huge amounts of data--matched the overall 90 percent accuracy of expert human imagery analysts in locating the missile sites.

SoftBank Group Corp. said it has upgraded the customer service capabilities of its Pepper humanoid robot, aiming to create demand in the business sector as Japan struggles with manpower shortages. The upgraded Pepper can take orders in English and Chinese, a new feature mainly designed to attract the restaurant industry, which is expected to see more foreign customers thanks to an increase in visitors from abroad, SoftBank said Monday. Customers can select a language on a screen located on Pepper's chest. The robot can recommend the day's special in the customer's preferred language. Orders can be placed via Pepper's chest screen, the company said.

Japan has become the second country to grand legal rights to a robot. Not long after Saudi Arabia gave citizenship to Sophia, albeit discredited by some as a PR stunt, Tokyo's Shibuya ward are following suit by giving residency to an AI entity named Shibuya Mirai – even giving it a birth certificate. Mirai, Japanese for "future", is a chatbot on the Japanese messaging service, Line. Unlike Sophia which was designed by Hanson Robotics to look human, Mirai does not have a physical presence. He exists solely online and was created as part of a drive to make the local government more accessible.

Aiming to reduce the possibilities of signals failing, Indian Railways has undertaken remote condition monitoring of the system, a new approach for the national transporter, to predict failures through the effective use of Artificial Intelligence. The Signalling system is vital for safe train operations and the railways completely depend on the health of its signalling assets alongwith real time information. Currently, the railways follow a manual maintenance system and adopt find-and-fix methods rather than predict-and-prevent approach. "Now, we are introducing remote condition monitoring using non-intrusive sensors for continuous online monitoring of signals, track circuits, axle counters and their sub-systems of interlocking, power supply systems including the voltage and current levels, relays, timers," said a senior Railway Ministry official involved with the project. The system entails the collection of inputs on a pre-determined interval and sending this to a central location.

Aiming to reduce the possibilities of signals failing, Indian Railways has undertaken remote condition monitoring of the system, a new approach for the national transporter, to predict failures through the effective use of Artificial Intelligence. The Signalling system is vital for safe train operations and the railways completely depend on the health of its signalling assets alongwith real time information. Currently, the railways follow a manual maintenance system and adopt find-and-fix methods rather than predict-and-prevent approach. "Now, we are introducing remote condition monitoring using non-intrusive sensors for continuous online monitoring of signals, track circuits, axle counters and their sub-systems of interlocking, power supply systems including the voltage and current levels, relays, timers," said a senior Railway Ministry official involved with the project. The system entails the collection of inputs on a pre-determined interval and sending this to a central location.

I joined Booking.com as a data scientist about two and a half years back, straight after a 3 years consulting gig in Dubai. Moving from consulting to a pure data science role was a big shift in my career and in hindsight I'm very happy I made that choice. In fact, I was already impressed with the company during my interviews. What I liked the most was that I was interviewed by peers who were already in the same role, which allowed for many'quality' interactions during the process and reaffirmed the recruiter's claim that the company had a'flat hierarchy'. Also the background of the interviewers was very diverse and interesting -- one had a PhD in astronomy, and the other was CTO of his own startup.