SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.05234 · NeurIPS · 2024

QGFN: Controllable Greediness with Action Values

Doina Precup, Emmanuel Bengio, Ling Pan, Elaine Lau, Stephen Zhewen Lu

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 8 functions out of this paper's own repositories and ran 7 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
yunglau/QGFN canonical 6 of 7
yunglau/qgfn extension 1 of 1
FunctionStatusWhere it lives
aggregate_iqm Ran yunglau/QGFN/utils/metrics.py
code served (permissive licence) · get_code("42c82ab33c19b894")
get_groupby_value Ran yunglau/QGFN/utils/metrics.py
code served (permissive licence) · get_code("31cec085da1cc1c5")
make_sh_script Ran yunglau/QGFN/utils/runs.py
code served (permissive licence) · get_code("5dc95db4192ee1e6")
mean_confidence_interval Ran yunglau/QGFN/utils/metrics.py
code served (permissive licence) · get_code("40f2f8185e2b1e3f")
scheduler Ran yunglau/qgfn/src/gflownet/data/mix_iterator.py
code served (permissive licence) · get_code("d8821f0b8834c2ef")
sqlite_load Ran yunglau/QGFN/utils/loaders.py
code served (permissive licence) · get_code("6e3e278b5881d761")
try_to_load_df Ran yunglau/QGFN/utils/plotting.py
code served (permissive licence) · get_code("70d87c0cd02ff899")
rna_sqlite_load Not yet run yunglau/QGFN/utils/loaders.py
code served (permissive licence) · get_code("30dd48f7c0ea8071")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Generative Flow Networks (GFlowNets; GFNs) are a family of energy-based generative methods for combinatorial objects, capable of generating diverse and high-utility samples. However, consistently biasing GFNs towards producing high-utility samples is non-trivial. In this work, we leverage connections between GFNs and reinforcement learning (RL) and propose to combine the GFN policy with an action-value estimate, Q, to create greedier sampling policies which can be controlled by a mixing parameter. We show that several variants of the proposed method, QGFN, are able to improve on the number of high-reward samples generated in a variety of tasks without sacrificing diversity.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2402.05234")
get_code_for_paper("2402.05234")
have("2402.05234")

Connect an agent — have() is free.