Anisotropic smoothing for change-point regression data

TFM de inteligencia artificial para predecir la severidad en incendios forestales utilizando datos históricos y modelos predictivos avanzados.

Attention-based spatiotemporal ConvLSTM models for wildfire spread prediction. Pairwise and patchwise self-attention mechanisms capture local fire front dynamics and long-range spotting effects. Validated on simulated percolation data and VIIRS satellite observations (California 2012–2020).