SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2310.02710 · ICLR · 2024

Local Search GFlowNets

Kaist, Emmanuel Bengio, Sungsoo Ahn, Minsu Kim, Jinkyoo Park, Yun Taeyoung

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 17 functions out of this paper's own repositories and ran 9 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
dbsxodud-11/ls_gfn canonical 9 of 16
dbsxodud-11/ls-gfn — 0 of 1
FunctionStatusWhere it lives
batch Ran dbsxodud-11/ls_gfn/gflownet/utils.py
pointer only (licence: NONE) · get_code("1a1ebe33ee0285e0")
collate_states_scores Ran dbsxodud-11/ls_gfn/gflownet/policy.py
pointer only (licence: NONE) · get_code("e62c1b64d8aa2840")
dist_calibration_error Ran dbsxodud-11/ls_gfn/gflownet/evaluate.py
pointer only (licence: NONE) · get_code("198490929c547443")
get_unique_children_in_x Ran dbsxodud-11/ls_gfn/gflownet/guide.py
pointer only (licence: NONE) · get_code("2dfc42fc19e46728")
make_mlp Ran dbsxodud-11/ls_gfn/gflownet/network.py
pointer only (licence: NONE) · get_code("266dbf4340bef3f5")
make_nodesummary_gnn Ran dbsxodud-11/ls_gfn/gflownet/network.py
pointer only (licence: NONE) · get_code("7e8123dabbedb095")
multi_set_distance Ran dbsxodud-11/ls_gfn/gflownet/monitor.py
pointer only (licence: NONE) · get_code("6dd46344ca1baf43")
pack Ran dbsxodud-11/ls_gfn/gflownet/utils.py
pointer only (licence: NONE) · get_code("1c23da6bab23ab31")
tensor_to_np Ran dbsxodud-11/ls_gfn/gflownet/utils.py
pointer only (licence: NONE) · get_code("973c7b0697f6a623")
BaseTBGFlowNet Not yet run dbsxodud-11/ls-gfn/gflownet/GFNs/basegfn.py
pointer only (licence: NONE) · get_code("23d514ace7e9b0fb")
anderson_darling Not yet run dbsxodud-11/ls_gfn/gflownet/evaluate.py
pointer only (licence: NONE) · get_code("cfafc9ffa57d7acf")
collate_probs Not yet run dbsxodud-11/ls_gfn/gflownet/policy.py
pointer only (licence: NONE) · get_code("797497ec7c51c286")
guide_logp Not yet run dbsxodud-11/ls_gfn/gflownet/guide.py
pointer only (licence: NONE) · get_code("ac05e44262397b23")
guide_sample Not yet run dbsxodud-11/ls_gfn/gflownet/guide.py
pointer only (licence: NONE) · get_code("6b1b2290357fc6dc")
make_convnet Not yet run dbsxodud-11/ls_gfn/gflownet/network.py
pointer only (licence: NONE) · get_code("b3c76e81a3360fbd")
make_full_exp Not yet run dbsxodud-11/ls_gfn/gflownet/data.py
pointer only (licence: NONE) · get_code("07d2630a9cb727da")
unique_keep_order_filter_children Not yet run dbsxodud-11/ls_gfn/gflownet/policy.py
pointer only (licence: NONE) · get_code("f928c1f4da5deecd")

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) are amortized sampling methods that learn a distribution over discrete objects proportional to their rewards. GFlowNets exhibit a remarkable ability to generate diverse samples, yet occasionally struggle to consistently produce samples with high rewards due to over-exploration on wide sample space. This paper proposes to train GFlowNets with local search, which focuses on exploiting high-rewarded sample space to resolve this issue. Our main idea is to explore the local neighborhood via backtracking and reconstruction guided by backward and forward policies, respectively. This allows biasing the samples toward high-reward solutions, which is not possible for a typical GFlowNet solution generation scheme, which uses the forward policy to generate the solution from scratch. Extensive experiments demonstrate a remarkable performance improvement in several biochemical tasks. Source code is available: https://github.com/dbsxodud-11/ls_gfn.

For agents

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

get_harvested_code_for_paper("2310.02710")
get_code_for_paper("2310.02710")
have("2310.02710")

Connect an agent — have() is free.