Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 7
Encontrada(s) 1 página(s)

Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

Remote sensing AI for crop planting in wildfire fuel mapping

  • Sánchez Gayet, Paula|||0009-0003-6372-716X
  • González Fernández, Irene|||0000-0001-8602-6495
  • Carrillo, Carlos|||0000-0003-3606-7517
  • Cortés Fité, Ana|||0000-0003-1697-1293
  • Suppi, Remo|||0000-0002-0373-8292
Accurate wildfire prediction requires updated, high-resolution fuel maps that account for seasonal vegetation variations. The flammability of crops varies by season, affecting the behavior of wildfires. This study combines remote sensing indices and machine learning to dynamically update fuel models in cropland zones. Using Sentinel-2 data, the status of the cropland is classified as "planted" or "unplanted", achieving 80% accuracy. Applied to a 2019 wildfire in Catalonia (Spain), the updated fuel map closely matched the observed fire spread. The methodology outperforms traditional approaches and is efficient, allowing for real-time updates based on seasonal changes.
Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

Assessing Physics Parameterizations Using Evolutionary Computation

  • Guerrero Calzas, Iciar
  • Rossetto, Lorenzo
  • Cortés Fité, Ana|||0000-0003-1697-1293
  • Hanzich, Mauricio|||0000-0003-4460-1118
  • Miró, Josep Ramon|||0000-0003-2838-6083
Hailstorms are intense, localized weather phenomena that can severely impact agriculture, infrastructure, and property, making precise forecasting essential for risk management. The Weather Research and Forecasting (WRF) model is widely used for numerical weather prediction, offering numerous physical parameterization options to represent atmospheric processes. However, due to the large number of possible configurations, identifying the most suitable configuration is a challenge. This research uses a genetic algorithm (GA) to systematically refine WRF physics schemes for hail prediction in Central Europe, specifically for the hail events of June 2022. Within this framework, WRF configurations are treated as individuals in a population that evolves through selection, crossover, and mutation over multiple iterations. Fitness is evaluated using the F2 score. This methodology allows to evaluate more than 2.4 million possible setups improving the WRF model's capacity to accurately represent hailstorms. This strategy provides a robust framework for testing a wide range of setups, proving its value in refining parameterizations to better forecast impactful weather phenomena.
Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

Efficient Code Region Characterization Through Automatic Performance Counters Reduction Using Machine Learning Techniques

  • Harutyunyan Gevorgyan, Suren|||0000-0003-2224-5730
  • César Galobardes, Eduardo|||0000-0002-9729-8557
  • Sikora, Anna|||0000-0003-0090-4109
  • Filipovič, Jiří|||0000-0002-5703-9673
  • Dutta, Akash|||0009-0007-0947-1182
  • Jannesari, Ali|||0000-0001-8672-5317
  • Alcaraz, Jordi|||0000-0002-9640-6763
Leveraging hardware performance counters provides valuable insights into system resource utilization, aiding performance analysis and tuning for parallel applications. The available counters vary with architecture and are collected at execution time. Their abundance and the limited number of registers for measurement make gathering laborious and costly. Efficient characterization of parallel regions necessitates a dimension reduction strategy. While recent efforts have focused on manually reducing the number of counters for specific architectures, this paper introduces a novel approach: an automatic dimension reduction technique for efficiently characterizing parallel code regions across diverse architectures. The methodology is based on Machine Learning ensembles because of their precision and ability at capturing different relationships between the input features and the target variables. Evaluation results show that ensembles can successfully reduce the number of hardware performance counters that characterize a code region. We validate our approach on CPUs using a comprehensive dataset of OpenMP regions, showing that any region can be accurately characterized by 8 relevant hardware performance counters. In addition, we also apply the proposed methodology on GPUs using a reduced set of kernels, demonstrating its effectiveness across various hardware configurations and workloads.
Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

Performance Optimization using Multimodal Modeling and Heterogeneous GNN

  • Dutta, Akash|||0009-0007-0947-1182
  • Alcaraz, Jordi|||0000-0002-9640-6763
  • Tehranijamsaz, Ali|||0009-0001-3678-5730
  • César Galobardes, Eduardo|||0000-0002-9729-8557
  • Sikora, Anna|||0000-0003-0090-4109
  • Jannesari, Ali|||0000-0001-8672-5317
Growing heterogeneity and configurability in HPC architectures has made auto-tuning applications and runtime parameters on these systems very complex. Users are presented with a multitude of options to configure parameters. In addition to application specific solutions, a common approach is to use general purpose search strategies, which often might not identify the best configurations or their time to convergence is a significant barrier. There is, thus, a need for a general purpose and efficient tuning approach that can be easily scaled and adapted to various tuning tasks. We propose a technique for tuning parallel code regions that is general enough to be adapted to multiple tasks. In this paper, we analyze IR-based programming models to make task-specific performance optimizations. To this end, we propose the Multimodal Graph Neural Network and Autoencoder (MGA) tuner, a multimodal deep learning based approach that adapts Heterogeneous Graph Neural Networks and Denoising Autoencoders for modeling IR-based code representations that serve as separate modalities. This approach is used as part of our pipeline to model a syntax, semantics, and structure-aware IR-based code representation for tuning parallel code regions/kernels. We extensively experiment on OpenMP and OpenCL code regions/kernels obtained from PolyBench, Rodinia, STREAM, DataRaceBench, AMD SDK, NPB, NVIDIA SDK, Parboil, SHOC, LULESH, XSBench, RSBench, miniFE, miniAMR, and Quicksilver benchmarks and applications. We apply our multimodal learning techniques to the tasks of (i) optimizing the number of threads, scheduling policy and chunk size in OpenMP loops and, (ii) identifying the best device for heterogeneous device mapping of OpenCL kernels. Our experiments show that this multimodal learning based approach outperforms the state-of-the-art in almost all experiments.
Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

Transferability of UNet-Based Downscaling Model for High-Resolution Temperature Data Across Diverse Regions

  • Benitez Benavides, Marc
  • Rodriguez, Mirta
  • Panadero, Javier|||0000-0002-3793-3328
  • Dutta, Omjyoti
  • Margalef, Tomàs|||0000-0001-6384-7389
High-resolution atmospheric data is essential for understanding local atmospheric processes, however it is computationally expensive to achieve such high resolutions through physical models. Recently, deep learning techniques, particularly those used in Single Image Super-Resolution, have emerged as a promising approach for statistical downscaling. However, much of the existing research has focused on enhancing model performance within small geographical regions, with limited attention given to the transferability of these models to diverse areas outside of their training domain. This paper introduces a methodology that evaluates the ability of a UNet model to downscale daily 2-meter temperature data outside its training region. The proposed approach uses one-third of the Contiguous United States to train the model, and assesses its performance on unseen areas. Our experimental design deliberately tests both spatial and temporal generalization, demonstrating that relatively compact models can effectively transfer downscaling capabilities to new regions. This results in improvements across key performance metrics including Mean Absolute Error, Root Mean Square Error, and Peak Signal-to-Noise Ratio. Additionally, our approach significantly reduces computational costs while improving downscaling accuracy across diverse climatic and topographic conditions.
Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

Impact of Mixed-Precision, A Way to Accelerate Data-Driven Forest Fire Spread Systems

  • Carrillo, Carlos|||0000-0003-3606-7517
  • Margalef, Tomàs|||0000-0001-6384-7389
  • Espinosa, Antonio|||0000-0002-6460-3789
  • Cortés Fité, Ana|||0000-0003-1697-1293
Every year, forest fires burn thousands of hectares of forest around the world and cause significant damage to the economy and people from the affected zone. For that reason, computational fire spread models arise as useful tools to minimize the impact of wildfires. It is well known that part of the forest fire forecast error comes from the uncertainty in the input data required by the models. Different strategies have been developed to reduce the impact of this input-data uncertainty during the last few years. One of these strategies consists of introducing a data-driven calibration stage where the input parameters are adjusted according to the actual evolution of the fire using an evolutionary scheme. In particular, the approach described in this work consists of a Genetic Algorithm (GA). This calibration strategy is computationally intensive and time-consuming. In the case of natural hazards, it is necessary to maintain a balance between accuracy and time needed to calibrate the input parameters. Most scientific codes have over-engineered the numerical precision required to obtain reliable results. In this paper, we propose to use a mixed-precision methodology to accelerate the calibration of the input parameters involved in forest fire spread prediction without sacrificing the accuracy of the forecast. The proposed scheme has been tested on a real fire. The results have led us to conclude that using the mixed-precision approach does not compromise the quality of the result provided by the forest fire spread simulator and can also speed up the whole evolutionary prediction system.
Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

A new paradigm for forest fire spread prediction, Faster decisions at high resolution

  • González Fernández, Irene|||0000-0001-8602-6495
  • Carrillo, Carlos|||0000-0003-3606-7517
  • Cortés Fité, Ana|||0000-0003-1697-1293
  • Espinosa, Antonio|||0000-0002-6460-3789
  • Margalef, Tomàs|||0000-0001-6384-7389
Climate change has led to a significant increase in the number of wildfire events and their severity. To mitigate their impact, it is necessary to be able to make quick decisions according to the fire behavior. In order to assist on these decisions, we rely on forest fire spread simulators. However, one of the main problems associated to these tools lies in finding the appropriate balance between execution time and simulation resolution, which has a direct impact on the accuracy of the simulated results. Simulations at high resolution are computational intensive in terms of execution time and, therefore, they are not used for real-time evolution forecast. FARSITE forest fire spread simulator has arisen as the most recognized simulation tool based on the Elliptical Wave Propagation (EWP) scheme. However, its main drawback consists of high execution times due the perimeter reconstruction algorithm used. In this paper, we propose an alternative algorithm to perform the reconstruction process based on the α -shapealgorithm. The proposed algorithm, Boundary α -shape algorithm, has been design bearing in mind that it could be easily parallelized. The obtained results when executing FARSITE by exchanging its reconstruction algorithm by the new one highlights that the Boundary α -shape allows a notable execution time reduction.