GitLab

In this work we have developed a procedure to predict fuel types for wildfires simulations. The procedure consists on the pre-processing of satellite imagery, its labeling and the prediction for each pixel with a neural network.

In this work we have compared and analyzed the uncertainties introduced by different fire-spotting parametrizations in an operational wildfire simulator. To do this, we have reproduced the effects of a real wildfire occurred in Italy, and we have app

FEniCS 2019 Presentation and preprint

This software library implements the federated decentralized version of the risk-based calibration algorithm for learning probabilistic classifiers such as naïve Bayes, quadratic discriminant analysis and logistic regression