GFlowNet Foundations
Authors: Yoshua Bengio, Salem Lahlou, Tristan Deleu, Edward J. Hu, Mo Tiwari, Emmanuel Bengio
JMLR 2023 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | In this paper, we show a number of additional theoretical properties of GFlow Nets, including a new local and efficient training objective called detailed balance for the analogy with MCMC. ... This paper extends the theory of the original GFlow Net construction (Bengio et al., 2021) in several directions, including a new local training objective called detailed balance (for the analogy with the detailed balance condition of Monte-Carlo Markov chains) as well as formulations enabling the calculation of marginal probabilities (or free energies) for subsets of variables, more generally for subsets of larger sets, or subgraphs, and their application to estimating entropy and mutual information. |
| Researcher Affiliation | Collaboration | Yoshua Bengio EMAIL Mila, Universi e de Montr eal, CIFAR, IVADO. Salem Lahlou EMAIL Mila, Universi e de Montr eal. Tristan Deleu EMAIL Mila, Universi e de Montr eal. Edward J. Hu EMAIL Mila, Universi e de Montr eal, Microsoft. Mo Tiwari EMAIL Stanford University. Emmanuel Bengio EMAIL Mila, Mc Gill University. |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. Methods are described mathematically and in prose. |
| Open Source Code | No | The paper does not provide any concrete access to source code for the methodology described. No links to repositories, explicit code release statements, or mention of code in supplementary materials are present. |
| Open Datasets | No | This paper is theoretical and focuses on the mathematical foundations and properties of GFlow Nets. It does not present experimental results or use any specific datasets, and therefore does not provide access information for publicly available or open datasets. |
| Dataset Splits | No | The paper is theoretical and does not present experimental results using datasets, thus no information about training/test/validation dataset splits is provided. |
| Hardware Specification | No | The paper is theoretical and does not describe any experiments that would require specific hardware. Therefore, no hardware specifications are mentioned. |
| Software Dependencies | No | The paper is theoretical and does not present any implementation details or experiments requiring specific software dependencies with version numbers. |
| Experiment Setup | No | The paper is theoretical and focuses on mathematical concepts and properties. It does not describe any empirical experiments or their setup, therefore, no specific experimental setup details, including hyperparameters or training configurations, are provided. |