No abstract

No abstract

No abstract

Summary Learning to rank (L2R) works by constructing a ranking model from training data so that, given a new query, the model is able to generate an effective rank of the objects for the query. Almost all work in L2R focus on ranking accuracy leaving performance and scalability overlooked. However, performance is a critical factor, especially when dealing with on‐demand queries. In this scenario, Learning to Rank using association rules has been shown to be extremely effective but only at a high computational cost. In this work, we show how to exploit parallelism on rule‐based systems to: i) drastically reduce L2R training datasets using selective sampling and ii) to generate query customized ranking models on the fly. We present parallel algorithms and GPU implementations for these two tasks showing that dataset reduction takes only a few seconds with speedups up to 148x over a serial baseline, and that queries can be processed in only a few milliseconds with speedups of 1000x over a serial baseline and 29x over a parallel baseline for the best case. We also extend the implementations to work with multiple GPUs, further increasing the speedup over the baselines and showing the scalability of our proposed algorithms.

Over the last decades, several real-time smoke prediction systems have been developed worldwide for air quality forecast to support decision making to control and manage anthropogenic pollutantions and smoke impacts. In Portugal, smoke modelling, as well as air quality forecast, has been developed by the research group GEMAC at the University of Aveiro, Portugal. However, the current forecast system does not integrate wildfire emissions. The ability from modelling to predict the behaviour of fire smoke in rural areas is an effective way to improve the efficiency of air quality and to prevent public health consequences. From 1980 to 2017, 4.4 Mha of cumulative burned area in rural fires, accounting for roughly half of Portugal's continental area, causing damage to infrastructure and lives. Hence, the forecast of smoke emissions has become of vital importance. There is a wide variety of models available to simulate fire-smoke phenomena. Nonetheless, it is necessary to consider computational aspects, resources, and goals to choose a suitable model to fit the purpose. In the ongoing FIRESMOKE project, developed by the GEMAC research group from University of Aveiro in Portugal, and GMAI from the Center for Weather Forecasting and Climate Studies in Brazil, the meteorological weather forecast model BRAMS-SFIRE is implemented to be part of the new version 5.6.2 of BRAMS. BRAMS-SFIRE model was coupled to simulate a broad integration between the surface fire fluxes and the atmospheric environment and presented a good accuracy in terms of the physics of the atmosphere and fire interaction. This project aims to improve the SFIRE model to include the crown fire behaviour. The goal is the incorporation of some formulations of the “crown fire potential” by linking models of surface and crown fire behaviour from Scott and Reinhardt (2001) and injecting fire smoke into the chemistry module of BRAMS. These developments are part of a whole system for forecasting and monitoring forest fire smoke emissions that incorporates other anthropogenic and biogenic sources of air pollution, to provide a public access service of atmospheric scope, over the domain of continental Portugal.