Despite the development of numerous adaptive optimizers, tuning the learning rate of stochastic gradient methods remains a major roadblock to obtaining good practical performance in machine learning. Rather than changing the learning rate at each iteration, we propose an approach that automates the most common hand-tuning heuristic: use a constant learning rate until "progress stops," then drop. We design an explicit statistical test that determines when the dynamics of stochastic gradient descent reach a stationary distribution. This test can be performed easily during training, and when it fires, we decrease the learning rate by a constant multiplicative factor. Our experiments on several deep learning tasks demonstrate that this statistical adaptive stochastic approximation (SASA) method can automatically find good learning rate schedules and match the performance of hand-tuned methods using default settings of its parameters. The statistical testing helps to control the variance of this procedure and improves its robustness.

The pace of technological change is rendering many job activities -- and the skills they require -- obsolete. Research by McKinsey suggests that globally more than 50% of the workforce is at risk of losing their jobs to automation, and a survey by the World Economic Forum suggests that 42% of the core job skills required today will change substantially by 2022. In this landscape of constant disruption, individuals, companies, and governments are fighting to ensure they have the skills to remain competitive. To shed light on the global skills landscape, Coursera recently released the first edition of our Global Skills Index (GSI) report. As the world's largest platform for higher education, Coursera brings together 40 million learners around the world with over 3,000 courses from leading universities and companies.

In the digital-first world, the value of artificial intelligence (AI) is more evident than ever, and many CEOs and business leaders are witnessing the positive impact it's having on their organizations. So it's no surprise that enterprises plan to double their number of AI projects within the next year. But despite the clear advantages of AI, businesses are still struggling to find the right talent to successfully implement and fully utilize these technologies. What's more, the disparity between AI optimism across the C-suite, and trust at the employee level, adds yet another barrier. A recent study we conducted at EY of U.S. CEOs and business leaders shows that, while a majority (84%) of CEOs realize the value of AI and its importance to their company's success, nearly one in three (31%) view a lack of skilled talent as a top barrier to AI adoption.

Mr. Yiannouka is the CEO of the World Innovation Summit for Education (WISE), a global think tank of the Qatar Foundation. WISE is dedicated to enabling the future of education through innovation. Its activities encompass research, capacity-building programs, and advocacy. WISE flagship initiatives include an annual series of research publications, a biennial global summit dubbbed the'Davos of education', the WISE edTech Accelerator, the WISE Innovation Awards, and the WISE Words podcast. Prior to joining WISE in August 2012, Stavros was the Executive Vice-Dean of the Lee Kuan Yew School of Public Policy (LKY School) at the National University of Singapore.

In last week's recap, I recounted how Alec Ross, author and technology policy advisor, had speculated that robot design, development and manufacturing would be critical to future economic well-being worldwide. What I didn't share was some of the background he provided that calls into question whether United States is positioning itself to be a leader in the robotics industry and the impact robotics will have on U.S. manufacturing jobs going forward. John Hitch has authored two recent articles on the robotics industry--a July 17 article, Reconciling Robot-Induced Anxiety and Admiration and an August 14 article, Manufacturing Obscurity is a Fate Worse than the Robopocalypse. These articles provide detail on these two questions above, and I highly recommend you read them. Here is some of the information they present.

Brent is a PhD Candidate in the Department of Computer Science and a member of the Φ Lab and Insight Lab. He was previously the instructor for MMASc 9251A: Professional Computing for Applied Scientists and presently the Teaching Assistant for Unstructured Data. As of March 1st 2019, Brent will also be a Mitacs Accelerate Intern. This work is with the Parkwood Institute and IBM with the target of improving mental health resources for Canadian Veterans. His research interests are two-fold.

Deadline: September 30, 2019 00:00 (CEST) Apply here Eligibility: Start-ups using machine learning (ML), artificial intelligence (AI), blockchain or extended reality, registered in one of UNICEF's programme countries, and have a working, open source prototype (or you are willing to make it open-source) showing promising results The UNICEF Innovation Fund is looking to make up to 100K equity-free investments to provide early-stage (seed) finance to for-profit technology start-ups that have the potential to benefit humanity. UNICEF's Innovation Fund has been specifically designed to finance early stage, open-source technology that can benefit children. The core motivation of the Innovation Fund is to identify "clusters" or portfolios of initiatives around emerging technology – so that UNICEF can both share markets and also learn about and guide these technologies to benefit children. It invests in solutions clusters around $100 billion industries in frontier technology spaces, such as blockchain, virtual and augmented reality, machine learning, and artificial intelligence. If you've got a start-up using machine learning (ML), artificial intelligence (AI), blockchain or extended reality, registered in one of UNICEF's programme countries, and have a working, open source prototype (or you are willing to make it open-source) showing promising results, the UNICEF Innovation Fund is looking for you.

Artificial intelligence (AI) is rapidly changing the educational landscape around the world. The technology is helping to simplify administrative tasks so that teachers can focus on communicating with students instead of spending valuable time on manually grading exams and homework. This new, tech-oriented approach may also help educators to create customized learning materials based on traditional syllabuses, and foster a personalized approach to learning. While India is serious about investing in AI, it's still way behind China, which plans to create a $150 billion AI-based tech sector by 2030. However, the education industry on the Indian subcontinent is experiencing dynamic growth and is currently revolutionizing the approach to learning.

MIT graduate student Justin Swaney is applying machine learning to music production. Rhythmic flashes from a computer screen illuminate a dark room as sounds fill the air. The snare drum sample comes out crisp and clean by itself, but turns muddy in the mix, no matter how the levels are set. Welcome to the world of modern music-making -- and its discontents. Today's digital music producers face a common dilemma: how to mesh samples that may sound great on their own but do not necessarily fit into a song like they originally imagined.