Andrew Yan-Tak Ng (Chinese: 吴恩达; born 1976) is a Chinese American computer scientist. He is the chief scientist at Baidu Research in Silicon Valley. In addition, he is an adjunct professor (formerly associate professor) at Stanford University. Ng is also the co-founder and chairman of Coursera, an online education platform. Ng researches primarily in machine learning and deep learning.

In this paper, we focus on online representation learning in non-stationary environments which may require continuous adaptation of model architecture. We propose a novel online dictionary-learning (sparse-coding) framework which incorporates the addition and deletion of hidden units (dictionary elements), and is inspired by the adult neurogenesis phenomenon in the dentate gyrus of the hippocampus, known to be associated with improved cognitive function and adaptation to new environments. In the online learning setting, where new input instances arrive sequentially in batches, the neuronal-birth is implemented by adding new units with random initial weights (random dictionary elements); the number of new units is determined by the current performance (representation error) of the dictionary, higher error causing an increase in the birth rate. Neuronal-death is implemented by imposing l1/l2-regularization (group sparsity) on the dictionary within the block-coordinate descent optimization at each iteration of our online alternating minimization scheme, which iterates between the code and dictionary updates. Finally, hidden unit connectivity adaptation is facilitated by introducing sparsity in dictionary elements. Our empirical evaluation on several real-life datasets (images and language) as well as on synthetic data demonstrates that the proposed approach can considerably outperform the state-of-art fixed-size (nonadaptive) online sparse coding of Mairal et al. (2009) in the presence of nonstationary data. Moreover, we identify certain properties of the data (e.g., sparse inputs with nearly non-overlapping supports) and of the model (e.g., dictionary sparsity) associated with such improvements.

Indeed, assessment may be better than compared with a conventional test. Griffin's research has found that the tasks on his platform do not exhibit the between nation bias (or Differential Item Functioning) that questions on the standardised, international PISA assessment purportedly suffer from (Kreiner & Christensen, 2014). They are also robust to differences in background language (Vista, Care and Griffin, 2014). The fact that assessment takes a back seat here begs the following question. Of what real worth is the psychometric modelling in the background?

Here I offer a few off-the-beaten-path interesting problems that you won't find in textbooks, data science camps, or in college classes. These problems range from applied maths, to statistics and computer science, and are aimed at getting the novice interested in a few core subjects that most data scientists master. The problems are described in simple English and don't require math / stats / probability knowledge beyond high school level. My goal is to attract people interested in data science, but who are somewhat concerned by the depth and volume of (in my opinion) unnecessary mathematics included in many curricula. I believe that successful data science can be engineered and deployed by scientists coming from other disciplines, who do not necessarily have a deep analytical background yet are familiar with data.

In this work we study the quantitative relation between the recursive teaching dimension (RTD) and the VC dimension (VCD) of concept classes of finite sizes. The RTD of a concept class $\mathcal C \subseteq \{0, 1\}^n$, introduced by Zilles et al. (2011), is a combinatorial complexity measure characterized by the worst-case number of examples necessary to identify a concept in $\mathcal C$ according to the recursive teaching model. For any finite concept class $\mathcal C \subseteq \{0,1\}^n$ with $\mathrm{VCD}(\mathcal C)=d$, Simon & Zilles (2015) posed an open problem $\mathrm{RTD}(\mathcal C) = O(d)$, i.e., is RTD linearly upper bounded by VCD? Previously, the best known result is an exponential upper bound $\mathrm{RTD}(\mathcal C) = O(d \cdot 2^d)$, due to Chen et al. (2016). In this paper, we show a quadratic upper bound: $\mathrm{RTD}(\mathcal C) = O(d^2)$, much closer to an answer to the open problem. We also discuss the challenges in fully solving the problem.

Billionaire investor Mark Cuban offered a perhaps bleak prediction on the future of jobs in an interview Friday. Billionaire investor Mark Cuban offered a perhaps bleak prediction on the future of jobs in an interview Friday with Bloomberg's Cory Johnson at the NBA All-Star Technology Summit in New Orleans. Discussing the swiftly evolving nature of jobs due to automation, he noted that across a broad array of industries, robots will replace human workers. Prompted by Johnson, he then made a bold proclamation about the types of skills and majors that will dominate in his version of the future labor market. Johnson: So essentially what you're making the case for is education and job training for grown ups.

What emerging innovations should learning leaders keep on their radar? Chief Learning Officer magazine's Kellye Whitney discusses trends to keep in mind in her video blog, with artificial intelligence leading the pack. AI as a learning delivery method is on the rise, with multiple applications surfacing beyond the typical high tech use cases, and growing adoption by industries like healthcare, automotive, and finance. Recognizing AI's potential, Udacity recently partnered with IBM's AI unit Watson, Didi Chuxing, and Amazon's AI voice assistant Alexa to offer a nanodegree in developing artificial intelligence. IBM will assist with designing curricula, and Udacity is also working with companies like Google and BMW to hire graduates of the program.

Education stands to benefit from rapid developments in artificial intelligence. But historically, adaptive learning software has been programmed in a top-down fashion. It asks a question, and if the child provides a particular answer, a set of prompts or tips might be shared or a new (perhaps easier, perhaps more advanced) question might be asked. Whether this is or is not AI is up for debate; however, the development of such programs is labor intensive and generally only effective in providing practice for concepts taught in class. As the Innovation Leader at an international education company, I work with both educational leaders on leveraging digital learning ecosystems and with education technology companies to improve their products for use in our schools.

I met Aman Naimat, Senior Vice President of Technology at Demandbase at their headquarters in San Francisco on my recent visit to California for Social Media Strategies Summit. Aman is working on leveraging the latest developments in Artificial Intelligence (AI) and data science for marketing and sales platforms. On this podcast, episode 150, Aman and I discuss how AI functions in account-based marketing (ABM). Before working with Demandbase Aman was previously founder and CTO of Spiderbook, a data-driven sales engine for account-based targeting. Aman has been building CRM systems since he was 19 and was the architect for the Oracle CRM Applications.