Algorithm design is a laborious process and often requires many iterations of ideation and validation. In this paper, we explore automating algorithm design and present a method to learn an optimization algorithm, which we believe to be the first method that can automatically discover a better algorithm. We approach this problem from a reinforcement learning perspective and represent any particular optimization algorithm as a policy. We learn an optimization algorithm using guided policy search and demonstrate that the resulting algorithm outperforms existing hand-engineered algorithms in terms of convergence speed and/or the final objective value.

We consider the problem of learning multi-ridge functions of the form f(x) = g(Ax) from point evaluations of f. We assume that the function f is defined on an l_2-ball in R^d, g is twice continuously differentiable almost everywhere, and A \in R^{k \times d} is a rank k matrix, where k << d. We propose a randomized, polynomial-complexity sampling scheme for estimating such functions. Our theoretical developments leverage recent techniques from low rank matrix recovery, which enables us to derive a polynomial time estimator of the function f along with uniform approximation guarantees. We prove that our scheme can also be applied for learning functions of the form: f(x) = \sum_{i=1}^{k} g_i(a_i^T x), provided f satisfies certain smoothness conditions in a neighborhood around the origin. We also characterize the noise robustness of the scheme. Finally, we present numerical examples to illustrate the theoretical bounds in action.

I've been re-reading Frank Harrell's Regression Modelling Strategies, a must read for anyone who ever fits a regression model, although be prepared - depending on your background, you might get 30 pages in and suddenly become convinced you've been doing nearly everything wrong before, which can be disturbing. I wanted to evaluate three simple modelling strategies in dealing with data with many variables. Using data with 54 variables on 1,785 area units from New Zealand's 2013 census, I'm looking to predict median income on the basis of the other 53 variables. The features are all continuous and are variables like "mean number of bedrooms", "proportion of individuals with no religion" and "proportion of individuals who are smokers". None of these is exactly what I would use for real, but they serve the purpose of setting up a competition of strategies that I can test with a variety of model validation techniques.

Wearable technology company LifeBEAM has powered products for the Israeli air force and brands like Under Armour and Samsung. What does that mean, exactly? Co-founder Omri Yoffe says the new "hearables," as he refers to these types of headphones, come with the ability to self-learn as you workout. Sony came out with a 300 "artificially intelligent" set of headphones in 2010, the Sony MDR-NC300D Digital Noise Canceling Earbuds, but the difference with LifeBEAMS buds, called Vi, is they have built-in software that adapts to your body using biometrics such as temperature, heart rate and other data measured through inner ear motion. Vi also includes a workout coach and music to match your pace.

While the world has been fixated on following the soap opera of financial markets, a more profound and ubiquitous development has been taking place worldwide - the rapid development in artificial intelligence and the fourth industrial revolution, which we think will mark an endless wave of disruptions. We believe artificial intelligence (AI) is almost ready for wider adoption by businesses, in turn providing opportunities, but also risks for investors. With corporate longevity already on the decline - according to McKinsey, one in five listed companies in the US may not last beyond the next five years - the integration of AI into business applications will have significant investment implications in the years to come. Similar to how companies with no core assets could become leaders in their industries today, AI companies have the potential to become tomorrow's industry leaders. As noted by Mr Tom Goodwin of the French media group Havas, who would have imagined just a few years back that the world's largest taxi firm (Uber) would own no vehicles, the world's largest accommodation provider (Airbnb) would operate no rooms, while the world's most valuable media company (Facebook) would create no content?

Positive news on the AI front recently when the US and UK governments both started public debate on the future of artificial intelligence. While there has been a lot of coverage on AI in the press over the recent months, it is very encouraging to see the governments of two of the leading contributing countries to this technology start taking its development seriously enough to have open public debates on the issues. Robotics and Autonomous Systems (RAS) is one of the'Eight Great Technologies' identified by the UK Government in 2012. A national strategy for RAS innovation from a'RAS Special Interest Group' has been published by Innovate UK. "Robots are now beating humans at even the most complex games, like Go. Artificial intelligence will play an increasing role in our lives over the coming years. From navigation systems to medical treatments and from new manufacturing techniques to unmanned vehicles, new applications are rapidly being developed that involve robotic decision making. It is important that the UK is ready with the research, innovation and skills to be able to fully take advantage of the opportunities and manage any risks. The global market for the AI sector is expected to grow to 2-6 trillion by 2025."

Specifically, using PL/Python, one can bring in countless Python libraries to process data close to the database. Here I will talk about my efforts to bring in the functionality of PySAL, a spatial analytics library written in Python and developed largely by Serge Rey, et al. at Arizona State University. PySAL makes available robust exploratory spatial data analysis related to spatial cluster and outlier detection, hotspot detection, spatial regression, and much more. Besides the wrappers we wrote for PySAL, we have written classes for bringing in machine learning methods such as random forest, linear regression, support vector machines, and neural networks from scikit-learn and Tensorflow. This talk will specifically cover the challenges we encountered programming in the PL/Python environment, collaborations with some of the PySAL developers, and the power of having spatial statistics and machine learning capabilities baked right into a cloud database.

Most of you know that I'm a learning consultant by trade and I apply the science of learning to real-world learning and performance improvement projects for my clients. You may also have noticed that one of my side interests is artificial and augmented intelligence. At least, I used to think that this was a side interest, only tenuously connected to my "day job," until several different threads converged in my brain and got me thinking: What if the Singularity – meaning the emergence of a true "artificial" intelligence (AI) -- has already happened and most of us just haven't noticed? Here's a short chronology of how my perception started to shift from "this is kind of cool" to "this could change everything." First I need to give you a bit of a disclaimer here: This story is not intended to be a detailed chronology of scientific developments in the fields discussed.

Using an in vitro analogue of the decision-making process, we show that the system employs an energy management strategy, switching between a low- and high-use mode depending on the outcome of the decision.

Artificial intelligence has come out of the dark corners of science fiction and is ready to become ubiquitous in the real world, according to one of the most influential tech gurus of the internet era. Writing on the company's website, Tim O'Reilly, founder and CEO of O'Reilly Media, say humanity is "at the beginning of an explosion of intelligent software" which could change everything. O'Reilly says: "It's easy to get excited about recent progress but much more difficult to understand what lies ahead. Researchers have built some excellent, but very specialized, systems. Still, we're far from anything that looks like a general intelligence system. "We have programs that can play Go, drive a car, chat, manage investments, and identify pictures; we don't have programs that can do all of the thousands of things that humans do every day.