train

train(data, model, loss_fun, num_epochs=10000, patience=20, lr=0.01, weight_decay=0.0, verbose=True, logger=<function <lambda>>)

train an embedding model

Parameters:

• data – network data

• model – embedding auto-encoder model

• loss_fun – loss function to use with model (takes arguments model, data)

• num_epochs – number of training epochs

• patience – patience for early stopping

• lr – learining rate (default: 0.01)

• weight_decay – weight decay for optimizer (default: 0.0)

• verbose – if True, display training progress (default: True)

• logger – function that receives the training loss as input and is called after each epoch (does nothing by default)

Returns:

trained model

This function uses the Adam optimizer for training.