Stochastic momentum methods have been widely adopted in training deep neural networks. However, their theoretical analysis of convergence of the training objective and the generalization error for prediction is still under-explored. This paper aims to bridge the gap between practice and theory by analyzing the stochastic gradient (SG) method, and the stochastic momentum methods including two famous variants, i.e., the stochastic heavy-ball (SHB) method and the stochastic variant of Nesterov's accelerated gradient (SNAG) method. We propose a framework that unifies the three variants. We then derive the convergence rates of the norm of gradient for the non-convex optimization problem, and analyze the generalization performance through the uniform stability approach. Particularly, the convergence analysis of the training objective exhibits that SHB and SNAG have no advantage over SG. However, the stability analysis shows that the momentum term can improve the stability of the learned model and hence improve the generalization performance. These theoretical insights verify the common wisdom and are also corroborated by our empirical analysis on deep learning.

Level assessment for foreign language students is necessary for putting them in the right level group, furthermore, interviewing students is a very time-consuming task, so we propose to automate the evaluation of speaker fluency level by implementing machine learning techniques. This work presents an audio processing system capable of classifying the level of fluency of non-native English speakers using five different machine learning models. As a first step, we have built our own dataset, which consists of labeled audio conversations in English between people ranging in different fluency domains/classes (low, intermediate, high). We segment the audio conversations into 5s non-overlapped audio clips to perform feature extraction on them. We start by extracting Mel cepstral coefficients from the audios, selecting 20 coefficients is an appropriate quantity for our data. We thereafter extracted zero-crossing rate, root mean square energy and spectral flux features, proving that this improves model performance. Out of a total of 1424 audio segments, with 70% training data and 30% test data, one of our trained models (support vector machine) achieved a classification accuracy of 94.39%, whereas the other four models passed an 89% classification accuracy threshold.

Amongst the analysts like Gartner, Forrester and IDC, Artificial Intelligence dominates their predictions for the future. You may well ask, 'what is artificial intelligence?'. The answer to this question is as under: "Artificial Intelligence is when a machine starts behaving like a human being" Technologically, machines do not have any intelligence and they just obey the commands given by their masters. However, when machines start displaying natural intelligence, then that is called artificial intelligence. In technical words, 'any machine that judges its environment and takes action to optimize its chance of achieving its goals.

Among the scant books in my tiny rented room in San Francisco, I've kept a spine-worn copy of Romeo and Juliet. It's the one I read in my high school English class, the pages yellowed, the margins filled with scribbled notes. Since the play was written in the 1590s, Shakespeare's portrayal of the nature of love--irrational, all-consuming--has been told and retold in countless movie adaptations. I hold onto the book to revisit those insights, and also because I'm prone to nostalgic literary tendencies like keeping old books. I am also a personal tech writer in 2018. It's my job to keep tabs on how our rapidly shifting technology is shaping not only how we communicate, but how we empathize, trust, show affection.

Technologies like virtual reality and artificial intelligence are revolutionizing the learning and development space. Just like the other functions, corporate learning has gone through a lot of transformations. However, one thing that hasn't changed is the importance of the learning function. In fact, with rapid disruptions in the business environment, re-skilling and upskilling has gained even more popularity. To gain more insights on the changing needs and technologies in the learning space, People Matters interacted with Sanjay Bahl, CEO and MD, Centum Learning.

A two-foot (60 cm) tall robot is being used to teach young children in Chinese schools. Keeko the robot bears a startling resemblance to Eve from the Pixar film Wall-E and moves around completely independently via inbuilt sensors and cameras. It also comes fitted with a small screen to interact with the pupils and is being used by teachers to tell stories and present logic problems to the kindergarten students. More than 600 kindergartens across the country have been equipped with a Keeko bot and the makers of the round machine hope to expand into Greater China and Southeast Asia. Children watch a Keeko robot (pictured) at the Yiswind Institute of Multicultural Education in Beijing, where the intelligent machines are telling stories and challenging kids with logic problems.

English-speaking robots will be helping out in some 500 Japanese classrooms from next year as the country seeks to improve English skills among both children and teachers using artificial intelligence. The education ministry is planning a pilot project costing around ¥250 million ($227,000) to improve students' notoriously weak oral and written skills in the language, an official said. "AI robots already on the market have various functions. For example, they can check the pronunciation of each student's English, which is difficult for teachers to do," added the official in charge of international education, who asked not to be named. AI robots "are just one example of the trial, and we are planning other measures," such as using tablet apps and having online lessons with native speakers, he said.

SEATTLE – Marking another phase in his education agenda, Bill Gates is now taking a more targeted approach to help struggling U.S. schools. The Bill and Melinda Gates Foundation is now funding groups working directly with clusters of public schools in some of the most impoverished regions of the country. Many of those third-party groups already had relationships with the world's largest philanthropy, and some of the grants went straight to a school district and charter schools organization. The foundation on Tuesday announced the first round of nearly $100 million for 19 program initiatives for middle and high schools in poor communities across 13 states. Gates pledged $460 million over the next five years to fund networks of school programs that help low-income and minority students get to college.

Active learning identifies data points to label that are expected to be the most useful in improving a supervised model. Opportunistic active learning incorporates active learning into interactive tasks that constrain possible queries during interactions. Prior work has shown that opportunistic active learning can be used to improve grounding of natural language descriptions in an interactive object retrieval task. In this work, we use reinforcement learning for such an object retrieval task, to learn a policy that effectively trades off task completion with model improvement that would benefit future tasks.

Solving statistical learning problems often involves nonconvex optimization. Despite the empirical success of nonconvex statistical optimization methods, their global dynamics, especially convergence to the desirable local minima, remain less well understood in theory. In this paper, we propose a new analytic paradigm based on diffusion processes to characterize the global dynamics of nonconvex statistical optimization. As a concrete example, we study stochastic gradient descent (SGD) for the tensor decomposition formulation of independent component analysis. In particular, we cast different phases of SGD into diffusion processes, i.e., solutions to stochastic differential equations. Initialized from an unstable equilibrium, the global dynamics of SGD transit over three consecutive phases: (i) an unstable Ornstein-Uhlenbeck process slowly departing from the initialization, (ii) the solution to an ordinary differential equation, which quickly evolves towards the desirable local minimum, and (iii) a stable Ornstein-Uhlenbeck process oscillating around the desirable local minimum. Our proof techniques are based upon Stroock and Varadhan's weak convergence of Markov chains to diffusion processes, which are of independent interest.