Florence d'Alché-Buc
Also known as Florence d'Alche-Buc
Florence d’Alché-Buc studies kernel methods, structured prediction and reliable machine learning, with applications in biology, medicine and industry. Her research includes work on functional outputs and prediction of graphs, alongside interpretation of audio-classification networks. She served as co-program chair for NeurIPS 2019. Her coauthored papers include graph prediction using optimal-transport methods and kernel approaches that make learning and inference more efficient. In 2024 her group reported Any2Graph, a method for supervised graph prediction with an optimal-transport loss. These projects show a focus on learning outputs with internal structure and understanding what predictive systems use.
Timeline
-
2019
Conference leadership
She served as co-program chair at NeurIPS 2019. That year, she also worked on kernel autoencoders and learning with outputs beyond a single scalar.
-
2022
Graphs and audio interpretation
She coauthored ICML work on predicting graphs using fused Gromov-Wasserstein barycenters. She also coauthored a NeurIPS paper on interpreting audio networks with non-negative matrix factorization. These are two specific applications of structured learning and interpretation.
-
2024
Efficient structured prediction
She coauthored work on sketched kernel regression and acceleration of learning and inference for structured prediction. Her coauthored Any2Graph paper was accepted at NeurIPS 2024. The work uses an optimal-transport loss for graph prediction.
-
As of October 2026
Research scope
Her research includes operator-valued kernel methods, structured output prediction, complex data, reliable machine learning and dynamical-system modeling. Her research supports applications in bioinformatics, medical research and industry.
Sources and credits
Photo: perso.telecom-paristech.fr; Florence d'Alché-Buc profile photograph; creator not stated. License: Not stated by source. Source: perso.telecom-paristech.fr. Cropped to a square and resized. Ask us to remove or replace this photo.