Completed
NeVAE: A variational autoencoder for graphs
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
A Deep Generative Model for Molecular Graphs by Niloy Ganguly
Automatically move to the next video in the Classroom when playback concludes
- 1 Start
- 2 Designing Random Graph Models using Variational Autoencoders...
- 3 Discovering new, plausible drug-like molecules
- 4 Generative models for molecule design
- 5 Limitations of current models
- 6 NeVAE: A variational autoencoder for graphs
- 7 The probabilistic encoder
- 8 Encoder has desirable properties
- 9 The probabilistic decoder
- 10 Decoder guarantees structural properties
- 11 Training is permutation invariant
- 12 Training is efficient
- 13 Experimental setup
- 14 Smooth, meaningful space of molecules
- 15 Quantitative evaluation metrics
- 16 Competing methods
- 17 Validity of the discovered molecules
- 18 Novelty of the discovered molecules
- 19 Uniqueness of the discovered molecules
- 20 Predicting & optimizing for molecule properties
- 21 Property prediction Sparse Gaussian Process
- 22 Property maximization Bayesian Optimization
- 23 Conclusions
- 24 Thanks!
- 25 Q&A
