This repository is a Python implementation of the RAT-CC model.

1. Create virtual env: python -m venv venv;
2. Activate the virtual environment: source venv/bin/activate;
3. Install requirements: pip install -r requirements.txt.

The repository defines the rat_cc function that returns an instance of the RAT-CC model. The required input parameters are:

• shape: The shape of the original data;
• emb_size: The size of the embedding (the compressed representation);
• n_layers: The number of hidden layers for the recurrent autoencoder (note that the encoder and the dencoder have the same number of hidden layers);
• units: The dimensionality of the output space for hidden layers;
• l: The lambda value.

# Instantiate the model
model = rat_cc((1000, 200, 2), 80, 5, 50)

# Train the model with training dataset
model.fit(
  x_train,
  x_train,
  validation_split=0.2,
)

# Get predictions
reconstructed, embedding = model.predict(x_test)