Carbon sequestration Cutting greenhouse-gas emissions alone won't be enough to prevent sharp increases in global temperatures. We'll also need to remove vast amounts of carbon dioxide from the atmosphere, which not only would be incredibly expensive but would present us with the thorny problem of what to do with all that CO2. A growing number of startups are exploring ways of recycling carbon dioxide into products, including synthetic fuels, polymers, carbon fiber, and concrete. That's promising, but what we'll really need is a cheap way to permanently store the billions of tons of carbon dioxide that we might have to pull out of the atmosphere. Grid-scale energy storage Renewable energy sources like wind and solar are becoming cheap and more widely deployed, but they don't generate electricity when the sun's not shining or wind isn't blowing. That limits how much power these sources can supply, and how quickly we can move away from steady sources like coal and natural gas.

Google just announced the 10 startups that have been shortlisted for the second calls of its Launchpad Accelerator program in India. All of the startups on the list have used artificial intelligence and machine learning to formulate their products. Google just announced the second wave of startups selected for their Launch Accelerator program in India. The program kicks off today with a one week mentorship programme boot camp organised by Google in Bengaluru which will be followed by more classes in April and May to address more specific issues -- lasting a total of three months. Aside from guidance, Google will also provide support for AI and ML, cloud computing, developing user interfaces, using the Android platform, online presence, product strategy and marketing.

Self-organization can be broadly defined as the ability of a system to display ordered spatio-temporal patterns solely as the result of the interactions among the system components. Processes of this kind characterize both living and artificial systems, making self-organization a concept that is at the basis of several disciplines, from physics to biology to engineering. Placed at the frontiers between disciplines, Artificial Life (ALife) has heavily borrowed concepts and tools from the study of self-organization, providing mechanistic interpretations of life-like phenomena as well as useful constructivist approaches to artificial system design. Despite its broad usage within ALife, the concept of self-organization has been often excessively stretched or misinterpreted, calling for a clarification that could help with tracing the borders between what can and cannot be considered self-organization. In this review, we discuss the fundamental aspects of self-organization and list the main usages within three primary ALife domains, namely "soft" (mathematical/computational modeling), "hard" (physical robots), and "wet" (chemical/biological systems) ALife. Finally, we discuss the usefulness of self-organization within ALife studies, point to perspectives for future research, and list open questions.

Fast linear transforms are ubiquitous in machine learning, including the discrete Fourier transform, discrete cosine transform, and other structured transformations such as convolutions. All of these transforms can be represented by dense matrix-vector multiplication, yet each has a specialized and highly efficient (subquadratic) algorithm. We ask to what extent hand-crafting these algorithms and implementations is necessary, what structural priors they encode, and how much knowledge is required to automatically learn a fast algorithm for a provided structured transform. Motivated by a characterization of fast matrix-vector multiplication as products of sparse matrices, we introduce a parameterization of divide-and-conquer methods that is capable of representing a large class of transforms. This generic formulation can automatically learn an efficient algorithm for many important transforms; for example, it recovers the $O(N \log N)$ Cooley-Tukey FFT algorithm to machine precision, for dimensions $N$ up to $1024$. Furthermore, our method can be incorporated as a lightweight replacement of generic matrices in machine learning pipelines to learn efficient and compressible transformations. On a standard task of compressing a single hidden-layer network, our method exceeds the classification accuracy of unconstrained matrices on CIFAR-10 by 3.9 points---the first time a structured approach has done so---with 4X faster inference speed and 40X fewer parameters.

In recent years, a variety of machine learning methods, such as random forest, gradient boosting trees and neural networks have gained popularity and been widely adopted. These methods are often flexible enough to uncover complex relationships in high-dimensional data without strong assumptions on the underlying data structure. Off-the-shelf software is available to put these algorithms into production [Pedregosa et al. (2011), Abadi et al. (2016) and Paszke et al. (2017)]. However, in regression and forecasting tasks, many of the machine learning methods only provide a point estimate, without any additional information regarding the uncertainty of the target quantity. Understanding uncertainties are often crucial in fields such as financial markets and risk analysis [Diebold et al. (1997), Timmermann (2000)], population and demographic studies [Wilson and Bell (2007)], transportation and traffic analysis [Zhu and Laptev (2017), Rodrigues and Pereira (2018)] and energy forecasting [Hong et al. (2016)].

Data Analytics is no more the future. Many people believed that it is just a flash in the pan and will wear off in some time. That fact is settled, and big data and analytics are here to stay. It is estimated that the market will grow at a CAGR of 30 percent between 2017 and 2023. This speedy growth will propel the market size for Big Data to reach a size of $80 billion.

As with all emerging technological trends, some elements of artificial intelligence are hyped out of proportion, some elements are ahead of their time, and some even incite fear. However, there remains some truth beneath the hype, cycles and buzzwords. Advancements in AI stand to benefit the energy sector but come with own limitations and practical concerns. Currently, AI, Machine Learning, and their other counterparts Deep Learning, Reinforcement Learning etc, have seen wide coverage in a variety of industries. But what do all these terms mean?

Finding the best light-harvesting chemicals for use in solar cells can feel like searching for a needle in a haystack. Over the years, researchers have developed and tested thousands of different dyes and pigments to see how they absorb sunlight and convert it to electricity. Sorting through all of them requires an innovative approach. Now, thanks to a study that combines the power of supercomputing with data science and experimental methods, researchers at the U.S. Department of Energy's (DOE) Argonne National Laboratory and the University of Cambridge in England have developed a novel "design to device" approach to identify promising materials for dye-sensitized solar cells (DSSCs). DSSCs can be manufactured with low-cost, scalable techniques, allowing them to reach competitive performance-to-price ratios.

Reservoir Computing (RC) is a popular methodology for the efficient design of Recurrent Neural Networks (RNNs). Recently, the advantages of the RC approach have been extended to the context of multi-layered RNNs, with the introduction of the Deep Echo State Network (DeepESN) model. In this paper, we study the quality of state dynamics in progressively higher layers of DeepESNs, using tools from the areas of information theory and numerical analysis. Our experimental results on RC benchmark datasets reveal the fundamental role played by the strength of inter-reservoir connections to increasingly enrich the representations developed in higher layers. Our analysis also gives interesting insights into the possibility of effective exploitation of training algorithms based on stochastic gradient descent in the RC field.

We present a new algorithm ASEBO for conducting optimization of high-dimensional blackbox functions. ASEBO adapts to the geometry of the function and learns optimal sets of sensing directions, which are used to probe it, on-the-fly. It addresses the exploration-exploitation trade-off of blackbox optimization, where each single function query is expensive, by continuously learning the bias of the lower-dimensional model used to approximate gradients of smoothings of the function with compressed sensing and contextual bandits methods. To obtain this model, it uses techniques from the emerging theory of active subspaces in the novel ES blackbox optimization context. As a result, ASEBO learns the dynamically changing intrinsic dimensionality of the gradient space and adapts to the hardness of different stages of the optimization without external supervision. Consequently, it leads to more sample-efficient blackbox optimization than state-of-the-art algorithms. We provide rigorous theoretical justification of the effectiveness of our method. We also empirically evaluate it on the set of reinforcement learning policy optimization tasks as well as functions from the recently open-sourced Nevergrad library, demonstrating that it consistently learns optimal inputs with fewer queries to a blackbox function than other methods.