In late May at the Future of Go Summit in Wuzhen, China, DeepMind's Go-playing artificial intelligence program AlphaGo won a three-game match over grandmaster Ke Jie and continued its dominance over human players. The Chinese Weiqi Association awarded AlphaGo 9-dan professional status, the highest possible rank for Go players, and Google subsidiary DeepMind announced AlphaGo's retirement, assigning its project team to other AI projects. AlphaGo was one of the biggest AI breakthroughs in recent memory because it beat the best human players at a game much more complex than chess and other board games. DeepMind relied on deep learning to program AlphaGo to learn Go in a similar manner to a human. But what makes deep learning deep?

"Battlefield 1," published by Electronic Arts, is a first-person shooter video game that depicts a female Russian soldier with a shaved head. "Battlefield 1," published by Electronic Arts, is a first-person shooter video game that depicts a female Russian soldier with a shaved head. The Playa Vista team behind the World War I video game "Battlefield 1" has two goals: Create an epic, all-out experience and ensure that their choices are true to history. Workers scan textbooks and online resources as they decide what characters, weapons and battles to depict in the game. They ask for help from experts on the Allies and Central Powers.

WeRobotics Global has become a premier forum for social good robotics. The feedback featured below was unsolicited. On June 1, 2017, we convened our first, annual global event, bringing together 34 organizations to New York City (full list below) to shape the global agenda and future use of robotics in the social good sector. WeRobotics Global was kindly hosted by the Rockefeller Foundation, the first donor to support our efforts. They opened the event with welcome remarks and turned it over to Patrick Meier from WeRobotics who provided an overview of WeRobotics and the big picture context for social sector robotics.

IoT solutions exploit operational data generated by Internet-connected smart "things" for the purpose of gaining operational insight and producing "better outcomes" (for example, create new business models, eliminate unscheduled maintenance, etc.). The explosive proliferation of IoT solutions will result in an exponential growth in the volume of IoT data, precipitating significant Information Governance issues: who owns the IoT data, what are the rights/duties of IoT solutions adopters towards this data, how to deal with Data Residency and Data Privacy issues, how to manage the lifecycle of IoT data, etc.? In his session at @ThingsExpo, Bassam Zarkout, founder of IGnPower, explored these issues and pressed on organizations to waste little time in setting their sails and embarking on this Information Governance journey. Speaker Bio Bassam Zarkout is a technology executive with 24 years of experience in Information Governance and related Content Management domains. As CTO at Swiss-based RSD, he implemented his vision for an innovative IG platform (GLASS) based on the principle of separating the rules from the tools from the repositories of information, thus enabling in-place governance of Information Assets. In April 2016, Bassam founded IGnPower, a consulting practice focused on applying IG principles to IoT data.

Both the arts and sciences are dissatisfied with verbal communication. To induce wonder, artists may forego language. To obtain knowledge, scientists often leave language behind. In his aptly titled essay "Science Has Outgrown the Human Mind and Its Limited Capacities," Ahmed Alkhateeb, a molecular cancer biologist at Harvard Medical School, suggests outsourcing research to artificial intelligence because "human minds simply cannot reconstruct highly complex natural phenomena efficiently enough in the age of big data." The problem is that language is a tool for the gathering of knowledge and appreciation of beauty by the whole society. Both the arts and sciences are dissatisfied with verbal communication. Scientists understand that language is a crude tool incapable of conveying abstract ideas. And to induce wonder, artists may forego language.

The U.S. has a racial wealth gap problem. By one estimate, at current levels of wealth growth it would take 228 years for the average black family to catch up with levels of wealth among white families. Thomas Shapiro explains some of the surprising reasons parity remains so elusive in his book, Toxic Inequality: How America's Wealth Gap Destroys Mobility, Deepens the Racial Divide, and Threatens Our Future.

Iterative Hard Thresholding (IHT) is a class of projected gradient descent methods for optimizing sparsity-constrained minimization models, with the best known efficiency and scalability in practice. As far as we know, the existing IHT-style methods are designed for sparse minimization in primal form. It remains open to explore duality theory and algorithms in such a non-convex and NP-hard problem setting. In this paper, we bridge this gap by establishing a duality theory for sparsity-constrained minimization with $\ell_2$-regularized loss function and proposing an IHT-style algorithm for dual maximization. Our sparse duality theory provides a set of sufficient and necessary conditions under which the original NP-hard/non-convex problem can be equivalently solved in a dual formulation. The proposed dual IHT algorithm is a super-gradient method for maximizing the non-smooth dual objective. An interesting finding is that the sparse recovery performance of dual IHT is invariant to the Restricted Isometry Property (RIP), which is required by virtually all the existing primal IHT algorithms without sparsity relaxation. Moreover, a stochastic variant of dual IHT is proposed for large-scale stochastic optimization. Numerical results demonstrate the superiority of dual IHT algorithms to the state-of-the-art primal IHT-style algorithms in model estimation accuracy and computational efficiency.

Node-perturbation learning is a type of statistical gradient descent algorithm that can be applied to problems where the objective function is not explicitly formulated, including reinforcement learning. It estimates the gradient of an objective function by using the change in the object function in response to the perturbation. The value of the objective function for an unperturbed output is called a baseline. Cho et al. proposed node-perturbation learning with a noisy baseline. In this paper, we report on building the statistical mechanics of Cho's model and on deriving coupled differential equations of order parameters that depict learning dynamics. We also show how to derive the generalization error by solving the differential equations of order parameters. On the basis of the results, we show that Cho's results are also apply in general cases and show some general performances of Cho's model.

Deep learning is the state-of-the-art in fields such as visual object recognition and speech recognition. This learning uses a large number of layers, huge number of units, and connections. Therefore, overfitting is a serious problem. To avoid this problem, dropout learning is proposed. Dropout learning neglects some inputs and hidden units in the learning process with a probability, p, and then, the neglected inputs and hidden units are combined with the learned network to express the final output. We find that the process of combining the neglected hidden units with the learned network can be regarded as ensemble learning, so we analyze dropout learning from this point of view.

We develop the method of stochastic modified equations (SME), in which stochastic gradient algorithms are approximated in the weak sense by continuous-time stochastic differential equations. We exploit the continuous formulation together with optimal control theory to derive novel adaptive hyper-parameter adjustment policies. Our algorithms have competitive performance with the added benefit of being robust to varying models and datasets. This provides a general methodology for the analysis and design of stochastic gradient algorithms.