Why Pool When You Can Flow? Active Learning with GFlowNets

arXiv 25 Jun 2026 2 min read

AI Insight

This paper introduces BALD-GFlowNet, a new active learning framework that uses Generative Flow Networks to sample informative molecules for drug discovery without needing to evaluate entire large databases. Unlike traditional pool-based active learning methods that must score billions of candidate molecules, this approach generates samples directly proportional to their informativeness using the BALD (Bayesian Active Learning by Disagreement) metric. In virtual screening experiments, the method matched the performance of standard BALD while producing more structurally diverse molecules and achieving computational efficiency independent of database size.

Why it matters

This work addresses a major computational bottleneck in drug discovery, where screening billions of potential molecules is prohibitively expensive. By enabling scalable active learning for molecular discovery, this approach could accelerate the identification of promising drug candidates while exploring more diverse chemical space.

Confidence

5/10Peer-reviewedBiology

Understand the Science

Drug discovery Concept coming soon Generative model Concept coming soon Active learning Concept coming soon

arXiv:2509.00704v2 Announce Type: replace-cross
Abstract: The scalability of pool-based active learning is limited by the computational cost of evaluating large unlabeled datasets, a challenge that is particularly acute in virtual screening for drug discovery. While active learning strategies such as Bayesian Active Learning by Disagreement (BALD) prioritize informative samples, it remains computationally intensive when scaled to libraries containing billions samples. In this work, we introduce BALD-GFlowNet, a generative active learning framework that circumvents this issue. Our method leverages Generative Flow Networks (GFlowNets) to directly sample objects in proportion to the BALD reward. By replacing traditional pool-based acquisition with generative sampling, BALD-GFlowNet achieves scalability that is independent of the size of the unlabeled pool. In our virtual screening experiment, we show that BALD-GFlowNet achieves a performance comparable to that of standard BALD baseline while generating more structurally diverse molecules, offering a promising direction for efficient and scalable molecular discovery.

Source: Why Pool When You Can Flow? Active Learning with GFlowNets