NeurIPS 2022

Timezone: »

Direct Sampling for extreme weather generation

Jorge Luis Guevara Diaz · Bianca Zadrozny · Campbell Watson · Daniela Szwarcman · Debora Lima · Dilermando Queiroz · Leonardo Tizzei · Maria Garcia · Maysa Macedo · Priscilla Avegliano

in LatinX in AI »

Direct Sampling is an algorithm that can generate synthetic data using only one training image and a set of conditioning points. This algorithm implicitly learns the conditional distribution of the probable values the data could take given a set of conditioning points and the training image.This algorithm does not learn an internal state, like parametric Machine Learning algorithms, but instead, it contains a pattern-matching algorithm that implicitly learns such conditional distribution. Thus, it is a non-parametric Machine learning algorithm that resembles the KNN approach. In this work, we explore the application of Direct Sampling for generating extreme precipitation events, which are precipitation weather fields with out-of-sample precipitation values. To this end, we propose to conditioning Direct Sampling not only in the training image and the conditioning points but also in a set of control points and a return precipitation level map to guide the out-of-sample precipitation value generation. We validate our approach with statistical metrics and connectivity metrics.

[ Poster] [ Topia]

Author Information

Jorge Luis Guevara Diaz (IBM Research)

Bianca Zadrozny (IBM Research)

Campbell Watson (IBM Research)

I'm an atmospheric scientist at IBM Research where my research spans climate, weather and water. I was a postdoc at Yale University with Prof. Ron Smith, and completed a PhD at the University of Melbourne with Prof. Todd Lane. Currently leading AI for Climate initiatives with the Future of Climate at IBM Research.

Daniela Szwarcman (IBM-Research Brazil)

Debora Lima (IBM Research)

Dilermando Queiroz (IBM Research)

Leonardo Tizzei (IBM Reserach)

Maria Garcia (IBM Research)

Maysa Macedo (IBM Research)

Maysa Macedo is a Researcher in the Visual Analytics & Insights group at IBM Research Brazil. She spent three years as Postdoctoral Fellow in intravascular images at Heart Institute – University of Sao Paulo Medical School. She developed her doctorate at Institute of Mathematics and Statistics - University of Sao Paulo in vessel tracking involving Magnetic Resonance and Computed Tomography. She is Bachelor in Informatics at State University of Rio de Janeiro and she is interested in segmentation, feature extraction, shape analysis and machine learning. More specifically, her current work analyzes tissues in atherosclerotic plaques from intracoronary optical coherence tomography using spatial-frequency analysis. Nowadays, she contributes as a reviewer for Journal of Health Informatics, Computer Methods and Programs in Biomedicine, Journal of Medical Imaging and Health Informatics and Research on Biomedical Engineering. Her research interests include medical imaging analysis, computer vision, machine learning and healthcare informatics.