This paper studies whether rationality can be computed. Rationality is defined as the use of complete information, which is processed with a perfect biological or physical brain, in an optimized fashion. To compute rationality one needs to quantify how complete is the information, how perfect is the physical or biological brain and how optimized is the entire decision making system. The rationality of a model (i.e. physical or biological brain) is measured by the expected accuracy of the model. The rationality of the optimization procedure is measured as the ratio of the achieved objective (i.e. utility) to the global objective. The overall rationality of a decision is measured as the product of the rationality of the model and the rationality of the optimization procedure. The conclusion reached is that rationality can be computed for convex optimization problems.

Using machine learning and the IoT, tractor fleet owners will be able to view and manage fleet utilization and predictive maintenance as well as forecast future tractor utilizations based on history, real-time weather and remote sensing satellite data. In sub-Saharan Africa, more than 60 percent of crops are managed manually, with less than 20 percent managed by tractors and other machinery, an unsustainable model as food demand increases due to population growth averaging 11 million per year. In addition, up to 50 percent of farmers suffer post-harvest losses annually due to poor planting practices. This was the motivation behind startup Hello Tractor, a mobile platform that enables farmers to access tractor services on demand. Using a mobile app, the service aggregates tractor service requests and pairs them with recommended tractors and operators, while simultaneously tracking how many hours each piece of equipment is in the field and area serviced.

Although science and the public celebrated the amazing predictive power of the new machine learningmethods, many researchers are left with some unease, simply because prediction does not imply causation. The ability to uncover causal relations is, however, at the core of most questions concerning the effects of particular policies, medical treatments, marketing campaigns, businessdecisions, etc. (see Athey, 2017, for a recent discussion). The recently rapidly expanding causal machine learning literature holds great promise for the improved estimation of causal effects by merging the statistics and econometrics literature oncausality with the supervised statistical and machine learning (ML) literature focussing on prediction. The classical causality literature clarifies the conditions needed for being able to estimate causal effects. It also shows how to transform a counterfactual causal problem into specific prediction problems (e.g., Imbens and Wooldridge, 2009). The latter literature on ML provides tools that can be highly effective in solving prediction problems (e.g.

Earlier this month, an intergovernmental conference in Marrakesh, Morocco brought together leaders from around the world to address global migration. After two days of deliberations, some 150 nations signed the Global Compact for Migration (GCM) agreement, which called for the implementation of more humane policies to ensure "safe, orderly and regular migration". But the conference and the agreement were very much a platform for Western doublespeak and hypocrisy. European countries and the United States are by far not concerned with the "safety" of migrants and refugees heading for their borders. In fact, the border security industry in Europe and the US is thriving and in both places, Israel with its infamous militarised security policies is serving as a role model and a major technology supplier.

Climbing teams spent over two weeks laser scanning Mount Rushmore in May 2010. High atop the Thomas Jefferson Memorial in Washington, D.C., is a layer of biofilm covering the dome, darkening and discoloring it. Biofilm is "a colony of microscopic organisms that adheres to stone surfaces," according to the U.S. National Park Service, which needed to get a handle on its magnitude to get an accurate cost estimate for the work to remove it. Enter CyArk, a non-profit organization that uses three-dimensional (3D) laser scanning and photogrammetry to digitally record and archive some of the world's most significant cultural artifacts and structures. CyArk spent a week covering "every inch" of the dome, processed the data, and returned a set of engineering drawings to the Park Service "to quantify down to the square inch how much biofilm is on the monument," says CEO John Ristevski.

Depression is ranked as the largest contributor to global disability and is also a major reason for suicide. Still, many individuals suffering from forms of depression are not treated for various reasons. Previous studies have shown that depression also has an effect on language usage and that many depressed individuals use social media platforms or the internet in general to get information or discuss their problems. This paper addresses the early detection of depression using machine learning models based on messages on a social platform. In particular, a convolutional neural network based on different word embeddings is evaluated and compared to a classification based on user-level linguistic metadata. An ensemble of both approaches is shown to achieve state-of-the-art results in a current early detection task. Furthermore, the currently popular ERDE score as metric for early detection systems is examined in detail and its drawbacks in the context of shared tasks are illustrated. A slightly modified metric is proposed and compared to the original score. Finally, a new word embedding was trained on a large corpus of the same domain as the described task and is evaluated as well.

Gradient-descent-based algorithms and their stochastic versions have widespread applications in machine learning and statistical inference. In this work we perform an analytic study of the performances of one of them, the Langevin algorithm, in the context of noisy high-dimensional inference. We employ the Langevin algorithm to sample the posterior probability measure for the spiked matrix-tensor model. The typical behaviour of this algorithm is described by a system of integro-differential equations that we call the Langevin state evolution, whose solution is compared with the one of the state evolution of approximate message passing (AMP). Our results show that, remarkably, the algorithmic threshold of the Langevin algorithm is sub-optimal with respect to the one given by AMP. We conjecture this phenomenon to be due to the residual glassiness present in that region of parameters. Finally we show how a landscape-annealing protocol, that uses the Langevin algorithm but violate the Bayes-optimality condition, can approach the performance of AMP.

Stagnation in the proportion of women in the workplace and women's declining representation in politics, coupled with greater inequality in access to health and education, offset improvements in wage equality and the number of women in professional positions, leaving the global gender gap only slightly reduced in 2018. This is according to the Forum's Global Gender Gap Report 2018, published today. According to the report, the world has closed 68% of its gender gap, as measured across four key pillars: economic opportunity; political empowerment; educational attainment; and health and survival. While only a marginal improvement on 2017, the move is nonetheless welcome as 2017 was the first year since the report was first published in 2006 that the gap between men and women widened. At the current rate of change, the data suggest that it will take 108 years to close the overall gender gap and 202 years to bring about parity in the workplace.

Understanding the loss surface of neural networks is essential for the design of models with predictable performance and their success in applications. Experimental results suggest that sufficiently deep and wide neural networks are not negatively impacted by suboptimal local minima. Despite recent progress, the reason for this outcome is not fully understood. Could deep networks have very few, if at all, suboptimal local optima? or could all of them be equally good? We provide a construction to show that suboptimal local minima (i.e. non-global ones), even though degenerate, exist for fully connected neural networks with sigmoid activation functions. The local minima obtained by our proposed construction belong to a connected set of local solutions that can be escaped from via a non-increasing path on the loss curve. For extremely wide neural networks with two hidden layers, we prove that every suboptimal local minimum belongs to such a connected set. This provides a partial explanation for the successful application of deep neural networks. In addition, we also characterize under what conditions the same construction leads to saddle points instead of local minima for deep neural networks.

Binary data matrices can represent many types of data such as social networks, votes or gene expression. In some cases, the analysis of binary matrices can be tackled with nonnegative matrix factorization (NMF), where the observed data matrix is approximated by the product of two smaller nonnegative matrices. In this context, probabilistic NMF assumes a generative model where the data is usually Bernoulli-distributed. Often, a link function is used to map the factorization to the $[0,1]$ range, ensuring a valid Bernoulli mean parameter. However, link functions have the potential disadvantage to lead to uninterpretable models. Mean-parameterized NMF, on the contrary, overcomes this problem. We propose a unified framework for Bayesian mean-parameterized nonnegative binary matrix factorization models (NBMF). We analyze three models which correspond to three possible constraints that respect the mean-parametrization without the need for link functions. Furthermore, we derive a novel collapsed Gibbs sampler and a collapsed variational algorithm to infer the posterior distribution of the factors. Next, we extend the proposed models to a nonparametric setting where the number of used latent dimensions is automatically driven by the observed data. We analyze the performance of our NBMF methods in multiple datasets for different tasks such as dictionary learning and prediction of missing data. Experiments show that our methods provide similar or superior results than the state of the art, while automatically detecting the number of relevant components.