My interview on the benefits of adaptive learning methods has been published in a white-book by Quantmetry.
The white-book, No. 5 “AI en production”, is free to download here.
Transcript of the interview (in french)
Qu’est-ce qu’une dérive ? L’apprentissage est souvent considéré comme une tâche statique. Cependant, en conditions réelles, les données évoluent constamment. C’est ce qu’on appelle une dérive conceptuelle. Par exemple, les marchés financiers sont instables : les risques de crédit ne sont pas les mêmes d’une année sur l’autre.
I have started a new potion as Research Fellow in the Machine Learning Group at the University of Waikato in New Zealand.
I have started a new potion as Postdoc at the Data, Intelligence and Graphs (DIG) group at Télécom ParisTech.
I have successfully defended my PhD thesis “Fast and Slow Machine Learning”.
Jury:\
M João Gama, University of Porto\
M Georges Hébrail, Électricité de France\
M Themis Palpanas, Université Paris Descartes\
M Ricard Gavaldà, Universitat Politècnica de Catalunya\
M Jesse Read, École Polytechnique\
M Albert Bifet, Télécom ParisTech (Directeur de recherche)\
M Talel Abdessalem, Télécom ParisTech (Directeur de thèse)
I am attending the 2018 IEEE International Conference on Big Data in Settle, USA, to present our paper
Learning Fast and Slow: A Unified Batch/Stream Framework..
Abstract: Data ubiquity highlights the need of efficient and adaptable data-driven solutions. In this paper, we present FAST AND SLOW LEARNING (FSL), a novel unified framework that sheds light on the symbiosis between batch and stream learning. FSL works by employing Fast (stream) and Slow (batch) Learners, emulating the mechanisms used by humans to make decisions.
I am attending the 2017 IEEE International Conference on Big Data in Boston, Massachusetts, to present our paper
Predicting over-indebtedness on batch and streaming data.
Abstract: Detecting over-indebtedness, the difficulties meeting household payment commitments, poses multiple Big Data challenges for banking institutions. We present a novel data-driven framework for predicting over-indebtedness on real-world data. A warning mechanism that generates predictions 6 months ahead, improving the chances of financial recovery. This framework is based on the combination of feature selection and supervised learning techniques, and uses data balancing for fine-tuning the predictive models.