A self-exciting event is something where one occurence of an event is more likely to trigger repeated occurences of that same event. For example, an event like an earthquake is likely to trigger succesive earthquakes (aftershocks) in the span of time immediately following the initial earthquake. As the time since the last earthquake increases, the probablility of another earthquake occuring declines. Events of this nature are known as Hawkes Processes.
The functions in this repository build and train a single-layer LSTM neural network to learn the elasped time between event occurences, known as inter-arrival times, in a Hawkes Process using statiscially simulated data for training.
The run_LSTM.lua script establishes default learning parameters and calls the train_hawkes_LSTM.lua to train the model.