An implementation of the Grammar VAE Paper and new architecture to improve results

If data does not exist, run:

• data\make_innovative_dataset.py

To train the model:

• python grammar_vae.py % the grammar model

attributes:

• --epochs % number of epochs
• --batch_size % batch size
• --latent_dim % size of latent space vector
• --data % data set name
• --model_name % model name
• --data_type % model name

Example:

• python grammar_vae.py --latent_dim=56 --epochs=50 % train a model with a 56D latent space and 50 epochs

To compute the model results:

Go to Theano-master and run:

• python setup.py install

The experiments with molecules require the rdkit library, which can be installed as described in http://www.rdkit.org/docs/Install.html.

1 - To generate the latent representations, go to:

equation_optimization/latent_features_and_targets_grammar/ or molecule_optimization/latent_features_and_targets_grammar/

and run:

• python generate_latent_features_and_targets.py

2 - To run all simulations go to: equation_optimization/ or molecule_optimization/ and run:

• python run_all_simulations.py

3 - Extract the final results by going to:

equation_optimization/ or molecule_optimization/

and run:

• python get_final_results.py