LinearAE latent embeddings should return z before the activation function is applied - Githubissues

ramanathanlab / mdlearn

Machine learning for molecular dynamics

MIT License

12 stars 7 forks source link

LinearAE latent embeddings should return z before the activation function is applied #42

Closed braceal closed 3 years ago

braceal commented 3 years ago

Otherwise the activation can 0 out negative embeddings.