Constrained Graph Variational Autoencoders for Molecule Design

Authors: Qi Liu, Miltiadis Allamanis, Marc Brockschmidt, Alexander Gaunt

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments compare our approach with a wide range of baselines on the molecule generation task and show that our method is successful at matching the statistics of the original dataset on semantically important metrics.
Researcher Affiliation Collaboration Qi Liu 1, Miltiadis Allamanis2, Marc Brockschmidt2, and Alexander L. Gaunt2 1Singapore University of Technology and Design 2Microsoft Research, Cambridge
Pseudocode No The paper contains Figure 1 which illustrates the generative procedure as a diagram, but there are no structured pseudocode or algorithm blocks.
Open Source Code Yes Our implementation of CGVAE can be found at https://github.com/Microsoft/ constrained-graph-variational-autoencoder.
Open Datasets Yes QM9 [26, 27], an enumeration of 134k stable organic molecules with up to 9 heavy atoms (carbon, oxygen, nitrogen and fluorine). ZINC dataset [12], a curated set of 250k commercially available drug-like chemical compounds. CEPDB [10, 11], a dataset of organic molecules with an emphasis on photo-voltaic applications.
Dataset Splits No The paper mentions training on datasets and sampling generated molecules, but does not specify train/validation/test dataset splits with percentages or counts for reproduction.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models or memory specifications used for running experiments.
Software Dependencies No The paper mentions using 'RDKit' but does not provide version numbers for RDKit or any other software dependencies.
Experiment Setup Yes Our experiments use S = 7. In our implementation, Eℓis a dimension-preserving linear transformation. C and Lℓare fully connected networks with a single hidden layer of 200 units and Re LU non-linearities. In our experiments, both g1 and g2 are implemented as linear transformations that project to scalars. We allow deviation from the pure VAE loss (λ1 = 1) following Yeung et al. [34].