Distribution-free uncertainty quantification for kernel methods by gradient perturbations

Machine Learning - We propose a data-driven approach to quantify the uncertainty of models constructed by kernel methods. Our approach minimizes the needed distributional assumptions, hence,...     

Teff (Eragrostis tef) processing,utilization and future opportunities: a review

The demand for food production has been constantly increasing due to rising population. In developed countries, for example, the emergence of regional production of old grains that are rarely utilized, along with the production of commonly consumed grains, has gained importance in recent years. These grains, known collectively as ancient or heirloom grains, have offered both farmers and consumers novel ways of cultivation and products with interesting taste, characteristics and nutritional value. Among the 30 000 plant species known, only five cereals currently provide more than 50% of the world's energy intake – bread wheat (Triticum aestivum), rice (Oryza sativa), sorghum (Sorghum bicolor), millets (Panicum sp.) and maize (Zea mays). The excessive utilization of these selected species has a great potential to cause genetic losses and difficulty in bridging future agricultural demands. Teff (Eragrostis tef), an ancient grain extensively cultivated in countries like Eritrea and Ethiopia, provides promising alternatives for new food uses since its nutritional value is significantly higher than most others cereal grains. The absence of gluten allows flexibility in food utilization since it can be directly substituted to gluten-containing products. The grain also offers an excellent balance of essential amino acids and minerals, which can fulfil the recommended daily intake and eliminates the need for fortification and enrichment. This review provides a general overview of the physical properties and nutritional composition of teff grains related to processing and applications in the food and feed industries. The current status of teff utilization, as well as the challenges in production and commercialization, and future opportunities is presented and discussed.     

Mixture ratio modeling of dynamic systems

Any knowledge extraction relies (possibly implicitly) on a hypothesis about the modelled-data dependence. The extracted knowledge ultimately serves to a decision-making (DM). DM always faces uncertainty and this makes probabilistic modelling adequate. The inspected black-box modeling deals with “universal” approximators of the relevant probabilistic model. Finite mixtures with components in the exponential family are often exploited. Their attractiveness stems from their flexibility, the cluster interpretability of components and the existence of algorithms for processing high-dimensional data streams. They are even used in dynamic cases with mutually dependent data records while regression and auto-regression mixture components serve to the dependence modeling. These dynamic models, however, mostly assume data-independent component weights, that is, memoryless transitions between dynamic mixture components. Such mixtures are not universal approximators of dynamic probabilistic models. Formally, this follows from the fact that the set of finite probabilistic mixtures is not closed with respect to the conditioning, which is the key estimation and predictive operation. The paper overcomes this drawback by using ratios of finite mixtures as universally approximating dynamic parametric models. The paper motivates them, elaborates their approximate Bayesian recursive estimation and reveals their application potential.     

When Does Communication Learning Need Hierarchical Multi-Agent Deep Reinforcement Learning

Abstract

Multi-agent systems need to communicate to coordinate a shared task. We show that a recurrent neural network (RNN) can learn a communication protocol for coordination, even if the actions to coordinate are performed steps after the communication phase. We show that a separation of tasks with different temporal scale is necessary for successful learning. We contribute a hierarchical deep reinforcement learning model for multi-agent systems that separates the communication and coordination task from the action picking through a hierarchical policy. We further on show, that a separation of concerns in communication is beneficial but not necessary. As a testbed, we propose the Dungeon Lever Game and we extend the Differentiable Inter-Agent Learning (DIAL) framework. We present and compare results from different model variations on the Dungeon Lever Game.     

Biodegradability Study of the Biosurfactant Contained in a Crude Extract from Corn Steep Water

Over the last few years, the global biosurfactant market has raised due to the increasing awareness among consumers, for the use of biological or bio-based products. Because of their composition, it can be speculated that these are more biocompatible and more biodegradable than their chemical homologous. However, at the moment, no studies exist in the literature about the biodegradability of biosurfactants. In this work, a biosurfactant contained in a crude extract, obtained from a corn wet-milling industry stream that ferments spontaneously in the presence of lactic acid bacteria, was subjected to a biodegradation study, without addition of external microbial biomass, under different conditions of temperature (5–45 °C), biodegradation time (15–55 days), and pH (5–7). For that, a Box–Behnken factorial design was applied, which allowed to predict the percentage of biodegradation for the biosurfactant contained in the crude extract, between the range of the independent variables selected in the study, obtaining biodegradation values between 3 and 80%. The percentage of biodegradation for the biosurfactant was calculated based on the increase in the surface tension of samples of the crude extract. Furthermore, it was also possible to predict the variation in t_1/2 for the biosurfactant (time to achieve the 50% of biodegradation) under different conditions.