Preprocess synthetic Hawkes data by padding the sequences. Args: loaded_hawkes_data: process_dim: tmax: Returns:

Compute the negative log-likelihood as a loss function. Args: seq_times: event occurrence timestamps seq_one

Implement `lrate` above

Read an event sequence and predict the next event time and type. Args: seq_times: # start from 0 seq_types:

Generate a simulated time interval sequences from original time interval sequences based on uniform distribution Args: interval_

helper: construct a models form hyper parameters

create a mask to hide padding and future input

Embedding labels to one-hot form. Produces an easy-to-use mask to select components of the intensity. Args: labels: class labels, sized [N

Update parameters and rate

mask out subsequent positions

:param vocab_size: total vocab size :param embed_size: embedding size of token embedding :param dropout: dropout rate

Apply residual connection to any sublayer with the same size.

Args: generator: tmax: end time for the simulations n_gen_seq: number of samples to take

Run predictions on testing dataset Args: seq_lengths: seq_types: seq_times: model: hmax: use_jupy