MolGAN: An implicit generative model for small molecular graphs. Cao, N. D. & Kipf, T. , 5, 2018. Paper abstract bibtex Deep generative models for graph-structured data offer a new angle on the problem of chemical synthesis: by optimizing differentiable models that directly generate molecular graphs, it is possible to side-step expensive search procedures in the discrete and vast space of chemical structures. We introduce MolGAN, an implicit, likelihood-free generative model for small molecular graphs that circumvents the need for expensive graph matching procedures or node ordering heuristics of previous likelihood-based methods. Our method adapts generative adversarial networks (GANs) to operate directly on graph-structured data. We combine our approach with a reinforcement learning objective to encourage the generation of molecules with specific desired chemical properties. In experiments on the QM9 chemical database, we demonstrate that our model is capable of generating close to 100% valid compounds. MolGAN compares favorably both to recent proposals that use string-based (SMILES) representations of molecules and to a likelihood-based method that directly generates graphs, albeit being susceptible to mode collapse.
@article{Cao-2018-ID320,
title = {Mol{GAN}: An implicit generative model for small molecular graphs},
abstract = {Deep generative models for graph-structured data offer a new angle on the
problem of chemical synthesis: by optimizing differentiable models that
directly generate molecular graphs, it is possible to side-step expensive
search procedures in the discrete and vast space of chemical structures. We
introduce Mol{GAN}, an implicit, likelihood-free generative model for small
molecular graphs that circumvents the need for expensive graph matching
procedures or node ordering heuristics of previous likelihood-based
methods. Our method adapts generative adversarial networks ({GAN}s) to
operate directly on graph-structured data. We combine our approach with a
reinforcement learning objective to encourage the generation of molecules
with specific desired chemical properties. In experiments on the {QM}9
chemical database, we demonstrate that our model is capable of generating
close to 100\% valid compounds. Mol{GAN} compares favorably both to recent
proposals that use string-based ({SMILES}) representations of molecules and
to a likelihood-based method that directly generates graphs, albeit being
susceptible to mode collapse.},
author = {Cao, Nicola De and Kipf, Thomas},
journal = {},
year = {2018},
month = {5},
url = {http://arxiv.org/abs/1805.11973v1},
url = {http://arxiv.org/pdf/1805.11973v1},
arxiv = {1805.11973v1},
keywords = {stat.{ML}},
file = {FULLTEXT:pdfs/000/000/000000320.pdf:PDF}
}
Downloads: 0
{"_id":"fNALqYzCdTfXWDs8v","bibbaseid":"cao-kipf-molgananimplicitgenerativemodelforsmallmoleculargraphs-2018","authorIDs":[],"author_short":["Cao, N. D.","Kipf, T."],"bibdata":{"bibtype":"article","type":"article","title":"MolGAN: An implicit generative model for small molecular graphs","abstract":"Deep generative models for graph-structured data offer a new angle on the problem of chemical synthesis: by optimizing differentiable models that directly generate molecular graphs, it is possible to side-step expensive search procedures in the discrete and vast space of chemical structures. We introduce MolGAN, an implicit, likelihood-free generative model for small molecular graphs that circumvents the need for expensive graph matching procedures or node ordering heuristics of previous likelihood-based methods. Our method adapts generative adversarial networks (GANs) to operate directly on graph-structured data. We combine our approach with a reinforcement learning objective to encourage the generation of molecules with specific desired chemical properties. In experiments on the QM9 chemical database, we demonstrate that our model is capable of generating close to 100% valid compounds. MolGAN compares favorably both to recent proposals that use string-based (SMILES) representations of molecules and to a likelihood-based method that directly generates graphs, albeit being susceptible to mode collapse.","author":[{"propositions":[],"lastnames":["Cao"],"firstnames":["Nicola","De"],"suffixes":[]},{"propositions":[],"lastnames":["Kipf"],"firstnames":["Thomas"],"suffixes":[]}],"journal":"","year":"2018","month":"5","url":"http://arxiv.org/pdf/1805.11973v1","arxiv":"1805.11973v1","keywords":"stat.ML","file":"FULLTEXT:pdfs/000/000/000000320.pdf:PDF","bibtex":"@article{Cao-2018-ID320,\n title = {Mol{GAN}: An implicit generative model for small molecular graphs},\n abstract = {Deep generative models for graph-structured data offer a new angle on the\n problem of chemical synthesis: by optimizing differentiable models that\n directly generate molecular graphs, it is possible to side-step expensive\n search procedures in the discrete and vast space of chemical structures. We\n introduce Mol{GAN}, an implicit, likelihood-free generative model for small\n molecular graphs that circumvents the need for expensive graph matching\n procedures or node ordering heuristics of previous likelihood-based\n methods. Our method adapts generative adversarial networks ({GAN}s) to\n operate directly on graph-structured data. We combine our approach with a\n reinforcement learning objective to encourage the generation of molecules\n with specific desired chemical properties. In experiments on the {QM}9\n chemical database, we demonstrate that our model is capable of generating\n close to 100\\% valid compounds. Mol{GAN} compares favorably both to recent\n proposals that use string-based ({SMILES}) representations of molecules and\n to a likelihood-based method that directly generates graphs, albeit being\n susceptible to mode collapse.},\n author = {Cao, Nicola De and Kipf, Thomas},\n journal = {},\n year = {2018},\n month = {5},\n url = {http://arxiv.org/abs/1805.11973v1},\n url = {http://arxiv.org/pdf/1805.11973v1},\n arxiv = {1805.11973v1},\n keywords = {stat.{ML}},\n file = {FULLTEXT:pdfs/000/000/000000320.pdf:PDF}\n}\n\n","author_short":["Cao, N. D.","Kipf, T."],"key":"Cao-2018-ID320","id":"Cao-2018-ID320","bibbaseid":"cao-kipf-molgananimplicitgenerativemodelforsmallmoleculargraphs-2018","role":"author","urls":{"Paper":"http://arxiv.org/pdf/1805.11973v1"},"keyword":["stat.ML"],"downloads":0,"html":"","metadata":{"authorlinks":{}}},"bibtype":"article","biburl":"http://woowoowoo.com/ideas/test86.bib","creationDate":"2021-02-10T01:28:56.734Z","downloads":0,"keywords":["stat.ml"],"search_terms":["molgan","implicit","generative","model","small","molecular","graphs","cao","kipf"],"title":"MolGAN: An implicit generative model for small molecular graphs","year":2018,"dataSources":["mkMPR3Lfz5vmkF7ur"]}