AI Insight
This paper introduces a new learning objective called "trajectory balance" for Generative Flow Networks (GFlowNets), which are methods for learning to generate complex objects like graphs or strings through sequences of actions. The authors demonstrate that their trajectory balance approach addresses inefficiencies in credit assignment across long action sequences that plagued previous methods (flow matching and detailed balance), which were analogous to temporal difference learning. Experiments across four domains show that trajectory balance improves convergence speed, sample diversity, and robustness when dealing with long action sequences and large action spaces.
Why it matters
This work has implications for machine learning applications requiring generation of complex compositional structures, such as molecular design, circuit design, or code generation. By improving the efficiency and reliability of GFlowNets, this method could accelerate scientific discovery and engineering tasks where exploring large discrete spaces is computationally challenging.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Generative flow networks (GFlowNets) are a method for learning a stochastic policy for generating compositional objects, such as graphs or strings, from a given unnormalized density by sequences of actions, where many possible action sequences may lead to the same object. We find previously proposed learning objectives for GFlowNets, flow matching and detailed balance, which are analogous to temporal difference learning, to be prone to inefficient credit propagation across long action sequences. We thus propose a new learning objective for GFlowNets, trajectory balance, as a more efficient alternative to previously used objectives. We prove that any global minimizer of the trajectory balance objective can define a policy that samples exactly from the target distribution. In experiments on four distinct domains, we empirically demonstrate the benefits of the trajectory balance objective for GFlowNet convergence, diversity of generated samples, and robustness to long action sequences and large action spaces.
Source: Trajectory balance: Improved credit assignment in GFlowNets