This section explains the design of DeepLog on a high level. DeepLog is a network that is implemented as a torch-train Module, which is an extension of torch.nn.Module including automatic methods to fit() and predict() data. This means it can be trained and used as any neural network module in the pytorch library.

In addition, we provide automatic methods to train and predict events given previous event sequences using the torch-train library. This follows a scikit-learn approach with fit(), predict() and fit_predict() methods. We refer to its documentation for a detailed description.