Attention Based Molecule Generation via Hierarchical Variational Autoencoder

arXiv:2402.16854v1 Announce Type: cross
Abstract: Molecule generation is a task made very difficult by the complex ways in which we represent molecules computationally. A common technique used in molecular generative modeling is to use SMILES strings with recurrent neural networks built into variational autoencoders - but these suffer from a myriad of issues: vanishing gradients, long-range forgetting, and invalid molecules. In this work, we show that by combining recurrent neural networks with convolutional networks in a hierarchical manner, we are …

abstract arxiv attention autoencoder autoencoders cs.lg generative generative modeling hierarchical modeling molecules networks neural networks q-bio.bm recurrent neural networks strings type variational autoencoders via