Despite widespread use of artificial intelligence, only 11% of companies say they see a significant financial return on their investment, according a new study by Boston Consulting Group in partnership with MIT Sloan Management Review. The low yield was revealed in a global survey of more than 3,000 managers with 57% saying they have piloted or deployed AI, a significant increase over three years ago. The authors of the study reported that companies can get the basics of AI right with the right data, technology, talent and strategy, but still see low ROI. "Only when organizations add the ability to learn with AI do significant benefits become likely," the authors said. The elements of learning with AI require companies to have a combination of machines learning autonomously, humans teaching machines and machines teaching humans.

As ENR FutureTech 2020 rolled into its fourth and final day, presentations offered a sense of emerging construction technologies and what comes next for developers and users amid a pandemic hit marketplace. Steve Jones, senior director of Dodge Data & Analytics, said that despite a significant decline in office and retail construction, there are bright spots for some non-residential market sectors, including healthcare and warehouses for e-commerce fulfillment. There are also hints of a shift toward renovation over new construction, which he described as a sign the construction economy is slowing. "We are in a cyclical business. This is just what it does," said Jones.

IT students in Denmark have created a software program that can determine the energy consumption and the amount of CO2 generated by the development of deep learning algorithms. According to their estimates, hardware used to train a deep learning algorithm can use worrying amounts of energy from an environmental standpoint. Whether browsing movies suggested by Netflix based on your viewing history, asking your voice assistant a question or interacting with a chatbot on an e-commerce website, all of these everyday online processes rely on deep learning algorithms. However, developing algorithms contributes to digital pollution. And it's precisely this environmental impact that students from the IT department of the University of Copenhagen have sought to quantify, using their Carbontracker software program.

Attention is increasingly turning to organic photovoltaic (OPV) solar cells after decades of relying on silicon, which is relatively expensive and lacks flexibility. OPV solar cells will be cheaper to make by using printing technologies, as well as being more versatile and easier to dispose of. But a major challenge is sorting through the huge volume of potentially suitable chemical compounds that can be synthesised (tailor-made by scientists) for use in OPVs. Researchers have tried using machine learning before to address this issue but many of those models were time consuming, required significant computer processing power and were difficult to replicate. Crucially, they did not provide enough guidance for the experimental scientists seeking to build new solar devices.

It's only as good as the methods and data it has been given. On its own, it doesn't know if information is missing, how much weight to give differing kinds of information or whether the data it draws on is incorrect or corrupted. It can't deal precisely with uncertainty or random events -- unless it learns how. Relying exclusively on data, as machine-learning models usually do, it does not leverage the knowledge experts have accumulated over years and physical models underpinning physical and chemical phenomena. It has been hard to teach the computer to organize and integrate information from widely different sources.

Reinforcement learning (RL) algorithms combined with modern function approximators such as kernel functions and deep neural networks have achieved significant empirical successes in large-scale application problems with a massive number of states. From a theoretical perspective, however, RL with functional approximation poses a fundamental challenge to developing algorithms with provable computational and statistical efficiency, due to the need to take into consideration both the exploration-exploitation tradeoff that is inherent in RL and the bias-variance tradeoff that is innate in statistical estimation. To address such a challenge, focusing on the episodic setting where the action-value functions are represented by a kernel function or over-parametrized neural network, we propose the first provable RL algorithm with both polynomial runtime and sample complexity, without additional assumptions on the data-generating model. In particular, for both the kernel and neural settings, we prove that an optimistic modification of the least-squares value iteration algorithm incurs an $\tilde{\mathcal{O}}(\delta_{\mathcal{F}} H^2 \sqrt{T})$ regret, where $\delta_{\mathcal{F}}$ characterizes the intrinsic complexity of the function class $\mathcal{F}$, $H$ is the length of each episode, and $T$ is the total number of episodes. Our regret bounds are independent of the number of states and therefore even allows it to diverge, which exhibits the benefit of function approximation.

Learning-based methodologies increasingly find applications in safety-critical domains like autonomous driving and medical robotics. Due to the rare nature of dangerous events, real-world testing is prohibitively expensive and unscalable. In this work, we employ a probabilistic approach to safety evaluation in simulation, where we are concerned with computing the probability of dangerous events. We develop a novel rare-event simulation method that combines exploration, exploitation, and optimization techniques to find failure modes and estimate their rate of occurrence. We provide rigorous guarantees for the performance of our method in terms of both statistical and computational efficiency. Finally, we demonstrate the efficacy of our approach on a variety of scenarios, illustrating its usefulness as a tool for rapid sensitivity analysis and model comparison that are essential to developing and testing safety-critical autonomous systems.

Deep reinforcement learning (DRL) has reached super human levels in complex tasks like game solving (Go and autonomous driving). However, it remains an open question whether DRL can reach human level in applications to financial problems and in particular in detecting pattern crisis and consequently dis-investing. In this paper, we present an innovative DRL framework consisting in two sub-networks fed respectively with portfolio strategies past performances and standard deviations as well as additional contextual features. The second sub network plays an important role as it captures dependencies with common financial indicators features like risk aversion, economic surprise index and correlations between assets that allows taking into account context based information. We compare different network architectures either using layers of convolutions to reduce network's complexity or LSTM block to capture time dependency and whether previous allocations is important in the modeling. We also use adversarial training to make the final model more robust. Results on test set show this approach substantially over-performs traditional portfolio optimization methods like Markowitz and is able to detect and anticipate crisis like the current Covid one.

Optimal Transport is a theory that allows to define geometrical notions of distance between probability distributions and to find correspondences, relationships, between sets of points. Many machine learning applications are derived from this theory, at the frontier between mathematics and optimization. This thesis proposes to study the complex scenario in which the different data belong to incomparable spaces. In particular we address the following questions: how to define and apply Optimal Transport between graphs, between structured data? How can it be adapted when the data are varied and not embedded in the same metric space? This thesis proposes a set of Optimal Transport tools for these different cases. An important part is notably devoted to the study of the Gromov-Wasserstein distance whose properties allow to define interesting transport problems on incomparable spaces. More broadly, we analyze the mathematical properties of the various proposed tools, we establish algorithmic solutions to compute them and we study their applicability in numerous machine learning scenarii which cover, in particular, classification, simplification, partitioning of structured data, as well as heterogeneous domain adaptation.

We develop a method to build distribution-free prediction intervals in batches for time-series based on conformal inference, called \Verb|EnbPI| that wraps around any ensemble estimator to construct sequential prediction intervals. \Verb|EnbPI| is closely related to the conformal prediction (CP) framework but does not require data exchangeability. Theoretically, these intervals attain finite-sample, approximately valid average coverage for broad classes of regression functions and time-series with strongly mixing stochastic errors. Computationally, \Verb|EnbPI| requires no training of multiple ensemble estimators; it efficiently operates around an already trained ensemble estimator. In general, \Verb|EnbPI| is easy to implement, scalable to producing arbitrarily many prediction intervals sequentially, and well-suited to a wide range of regression functions. We perform extensive simulations and real-data analyses to demonstrate its effectiveness.