Max Hodak

I don't think the Travis error here was related to my commit, see: https://travis-ci.org/scambra/devise_invitable/jobs/22266880

What's the origin of these? Are these measured logPs based on looking up the SMILES? Or cLogP from rdkit? Something else?

What would be really great is if you could alter the `smiles_50k.h5` file to include a `clogp` column with this data indexed to the right rows, instead of including this...

So I let a 3-LSTM run overnight on Saturday and the loss fell to near zero, but it was definitely overfitting; it clearly wasn't extracting any interesting information about the...

I'm using presence/absence from the set. SMILES validity is an arguably even easier metric, as `CCCCCcccccccccccccccccccccccccccccccccc` is valid SMILES but not representative of the distribution we want to learn.

This is pretty typical of attempts to train my network right now: Sampling from which gives me stuff that looks like `Caaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa`

See: https://github.com/maxhodak/keras-molecules/blob/gan/SMILES_GAN.ipynb

On pretraining, worth noting that if I don't pretrain the generator, no interesting training happens at all when I try and train the GAN. Discriminator loss just goes to 0...

I'm not sure that matters... this isn't an autoencoder; the input is just a source of entropy. The nonlinearities in the generator network should mean the distribution of the input...

I've got something looking much better now after working in a bunch of the tricks linked above, though it still has a lot of room for improvement: After 200 iterations...