Published in ICLR 2024 Tackling Climate Change with Machine Learning, 2024
Calibration of the neural networks provides a way to ensure our confidence in the predictions. We calibrate a UNet++ based architecture. We show that with a slight trade-off between prediction error and calibration error, it is possible to get more reliable and sharper forecasts.
Published in CHI 2024 Computational Methodologies for Understanding, Automating, and Evaluating User Interfaces, 2024
We propose a novel content-aware graph-based representation that can infer suggestions for color palettes and color assignments to designers by capturing the types of elements (e.g., text, background, decoration), the associated colors, and the relationships between UI elements. This allows dynamically suggesting color palette selection in re- sponse to changes in color preferences and assigning updated colors for UI elements by employing graph neural networks (GNNs).
Published in Environmental Data Sciences 2023, 2023
A UNet++ architecture with circular convolutions is trained using spatiotemporal climate data for sub-seasonal temperature forecasting. The effect of ancillary climate variables have been investigated. The model predicts significantly better than the mean CMIP6 ensemble between 2016 and 2021. Both models predict the summer months more accurately than the winter months.