Before running conditional generation experiment on Noist MNIST, it's necessary to pretrain classifier, that will be used to compute validation metrics.

To pretrain classifier, please run

$ python pretrain_mnist_clf.py

--gpu= - gpu for training, if -1 then will train on CPU

To train some generative model on Noisy MNIST dataset, please, run following command.

$ python mnist_condgen_experiment.py

--model= - can be biaae, uniaae, lat_saae, saae, cvae, jmvae, vib or vcca

--gpu= - gpu for training, if -1 then will train on CPU

Before running training of model, please, pretrain RNN encoder and decoder.

To pretrain them, please run

$ python pretrain_rnn_enc_dec.py

--gpu= - gpu for training, if -1 then will train on CPU

To train conditional generative model (generate molecule by given transcriptome change) on LINCS dataset, please, run following command.

$ python lincs_experiment.py

--model= - can be biaae, uniaae, lat_saae, saae, cvae, jmvae, vib or vcca

--gpu= - gpu for training, if -1 then will train on CPU

To train conditional generative model (generate transcriptome change by given molecule) on LINCS dataset, please, run following command.

$ python lincs_experiment_reverse.py

--model= - can be biaae, uniaae, lat_saae, saae, cvae, jmvae, vib or vcca

--gpu= - gpu for training, if -1 then will train on CPU