Kernel-Based Few-Shot Regression and Applications in Drug Discovery

Due to the significant costs of data generation, many prediction tasks within drug discovery are by nature few-shot regression (FSR) problems, including accurate modeling of biological assays. The first part of the talk will provide some background for understanding the challenges faced by current models in drug discovery and the second part will present new algorithms that we have developed to face these challenges.

Date
  • 28 février 2020
Heure

13h30 à 14h30

Localisation

Université Laval
Pavillon Adrien-Pouliot
Local 3370

Coûts

Conférence de Prudencio Tossou, étudiant au doctorat en informatique, membre du Groupe de recherche en apprentissage machine de l’Université Laval (GRAAL)

Due to the significant costs of data generation, many prediction tasks within drug discovery are by nature few-shot regression (FSR) problems, including accurate modeling of biological assays. The first part of the talk will provide some background for understanding the challenges faced by current models in drug discovery and the second part will present new algorithms that we have developed to face these challenges.

More precisely, we have developed Adaptive deep kernel learning, that consist of learning deep networks in combination with kernel functions using differentiable kernel algorithms. The algorithm learns to find the appropriate kernel for each task during inference. It thus performs more effectively with complex task distributions, outperforming current state-of-the-art algorithms on both toy and novel, real-world benchmarks that we introduced.

Restons en contact!

Vous souhaitez être informé des nouvelles et activités de l'IID? Abonnez-vous dès maintenant à notre infolettre mensuelle.