Good science fiction excels at tapping contemporary anxieties to forecast the future fate of humanity. Consider the list of workforce automation fears showcased in the recent "Kerblam" episode of the TV series Doctor Who: Robotized megacorporations, computer-controlled commerce, ubiquitous unemployment, human irrelevance and destructive despair. There's no spoiler in sharing that -- aside from some teleportation and a sonic screwdriver -- everything in the story line is already pretty plausible. The Doctor may be sci-fi fantasy, but the issues are real. Artificial intelligence (AI) technology is already reshaping many manufacturing and service industries, and it is rapidly disrupting other sectors as well -- for both good and ill.

When we talk about artificial intelligence, there is hardwired imagery of massive thinking machines working in a science-fiction environment. Since AI technology has become the talk among many scholars and researchers, it is essential for more students to know about its functions. After all, they are going to give creative shape towards the growth of AI's industry in the future. Hence, to understand the rapid advancement of technology and master the concept of AI, many educationists from across the world are initiating educational institutions to include Artificial Intelligence in their syllabus. Promoting just that, the Indian Institute of Technology (IIT) Hyderabad will launch a full-fledged B tech program in AI from the coming academic year 2019-20. And the admissions to the new course will be made through the Joint Entrance Examination (JEE) Advanced course.

VideoLectures.NET is part of the H2020 project MOVING, which has been working on developing new and more effective methods for lecture video fragmentation and fragment-level annotation, to allow for fine-grained access to lecture video collections. In the latest MOVING method, developed by CERTH (also a member, and coordinator of the MOVING consortium), automatically-generated speech transcripts of the lecture video are analysed with the help of word embeddings that are generated from pre-trained state-of-the-art neural networks. This lecture video fragmentation method is part of the MOVING platform, and its results are also being ingested in the VideoLectures.NET platform, making it possible for the users of both platforms to access and view specific fragments of lecture videos that cater to their information needs. For now, the fragments are accessible only for some lectures in VideoLectures.NET (testing phase); see for instance the lecture on deep learning. The fragments are presented as "chapters" to the right of the video player window, and can serve as a tool to find particular video parts easier and faster.

Not all these contributions were from 2018, but the few selected below were among the most visited in 2018. Some were heavily featured, so it does not mean that they represent the average DSC interest. A bigger list featuring 900 most popular articles can be found here. I am still working on categorizing them, and may hire an intern to work on this project, using material described in this article or this one. Likewise, one of my old academic papers published in 1994 in IEEE Pattern Analysis and Machine Intelligence, an obscure, very theoretical math paper, entitled "Simulated Annealing: a Proof of Convergence" is getting a lot of traction recently in AI circles - giving me an academic score better than many university professors.

A new breed of intelligent video surveillance is being installed in schools around the country -- tech that follows people around campus and detects unusual behaviors. Axios' Kaveh Waddell reports: This new phase in campus surveillance responds to high-profile school shootings like the one in Parkland, Florida, last February. School administrators are now reaching for security tech that keeps a constant, increasingly sophisticated eye on halls and classrooms. Background: Schools are experimenting wildly with technology in order to secure students, deploying facial recognition, license plate readers, microphones for gunshot detection and even patrol robots. The tech the district wants is benignly branded "intelligent video analytics."

Surrogate modeling and uncertainty quantification tasks for PDE systems are most often considered as supervised learning problems where input and output data pairs are used for training. The construction of such emulators is by definition a small data problem which poses challenges to deep learning approaches that have been developed to operate in the big data regime. Even in cases where such models have been shown to have good predictive capability in high dimensions, they fail to address constraints in the data implied by the PDE model. This paper provides a methodology that incorporates the governing equations of the physical model in the loss/likelihood functions. The resulting physics-constrained, deep learning models are trained without any labeled data (e.g. employing only input data) and provide comparable predictive responses with data-driven models while obeying the constraints of the problem at hand. This work employs a convolutional encoder-decoder neural network approach as well as a conditional flow-based generative model for the solution of PDEs, surrogate model construction, and uncertainty quantification tasks. The methodology is posed as a minimization problem of the reverse Kullback-Leibler (KL) divergence between the model predictive density and the reference conditional density, where the later is defined as the Boltzmann-Gibbs distribution at a given inverse temperature with the underlying potential relating to the PDE system of interest. The generalization capability of these models to out-of-distribution input is considered. Quantification and interpretation of the predictive uncertainty is provided for a number of problems.

We have shown previously that our parameter-reduced variants of Long Short-Term Memory (LSTM) Recurrent Neural Networks (RNN) are comparable in performance to the standard LSTM RNN on the MNIST dataset. In this study, we show that this is also the case for two diverse benchmark datasets, namely, the review sentiment IMDB and the 20 Newsgroup datasets. Specifically, we focus on two of the simplest variants, namely LSTM_6 (i.e., standard LSTM with three constant fixed gates) and LSTM_C6 (i.e., LSTM_6 with further reduced cell body input block). We demonstrate that these two aggressively reduced-parameter variants are competitive with the standard LSTM when hyper-parameters, e.g., learning parameter, number of hidden units and gate constants are set properly. These architectures enable speeding up training computations and hence, these networks would be more suitable for online training and inference onto portable devices with relatively limited computational resources.

Machine learning models have become more and more complex in order to better approximate complex functions. Although fruitful in many domains, the added complexity has come at the cost of model interpretability. The once popular k-nearest neighbors (kNN) approach, which finds and uses the most similar data for reasoning, has received much less attention in recent decades due to numerous problems when compared to other techniques. We show that many of these historical problems with kNN can be overcome, and our contribution has applications not only in machine learning but also in online learning, data synthesis, anomaly detection, model compression, and reinforcement learning, without sacrificing interpretability. We introduce a synthesis between kNN and information theory that we hope will provide a clear path towards models that are innately interpretable and auditable. Through this work we hope to gather interest in combining kNN with information theory as a promising path to fully auditable machine learning and artificial intelligence.

Modern enterprises are struggling to gain insights from an exploding number of database management systems and ever-growing data volumes. SQL Server 2019 can help you overcome the challenges of integrating data and bring AI and machine learning to all of your data, structured and unstructured. It can also help you better manage your relational data right now. In this webinar, Introduction to SQL Server 2019, hear from Debbi Lyons, Senior Product Marketing Manager, Travis Wright, Principal Program Manager, and Bob Ward, Principal Architect at Microsoft discuss the latest updates and features for the new SQL Server release, including introducing the new big data cluster with intelligence over any data, how SQL Server 2019 enhances the developer experience, and using tools including Azure Data Studio. Listen to the webinar on-demand to learn more about what's new in SQL Server 2019, including how to: With SQL Server 2019 big data clusters, Apache SparkTM and HDFS are packaged together with SQL Server as a single, integrated solution.

What inspired you to take an interest in robots? I've always played with robots. For example, I remember a competition in high school where my friends and I built a robotic arm to move the chess pieces on the chessboard. It seems very trivial now, but way back then, the robots were all primitive and as high school students, we thought that building a robot that could do that was a big deal. Graduate students Ashutosh Saxena, left, and Morgan Quigley, center, and Ng were part of a large effort to develop a robot to see an unfamiliar object and ascertain the best spot to grasp it.