Fully Neural Network based Model for General Temporal Point Processes
This work addresses the need for more flexible and expressive models in temporal point processes, which are used in various applications like event prediction, but it is incremental as it builds on existing RNN-based approaches.
The authors tackled the limitation of existing RNN-based temporal point process models, which assume specific functional forms for intensity functions, by proposing a fully neural network-based model that represents the intensity function in a general manner using a feedforward network and its derivative, achieving competitive or superior performance on synthetic and real datasets.
A temporal point process is a mathematical model for a time series of discrete events, which covers various applications. Recently, recurrent neural network (RNN) based models have been developed for point processes and have been found effective. RNN based models usually assume a specific functional form for the time course of the intensity function of a point process (e.g., exponentially decreasing or increasing with the time since the most recent event). However, such an assumption can restrict the expressive power of the model. We herein propose a novel RNN based model in which the time course of the intensity function is represented in a general manner. In our approach, we first model the integral of the intensity function using a feedforward neural network and then obtain the intensity function as its derivative. This approach enables us to both obtain a flexible model of the intensity function and exactly evaluate the log-likelihood function, which contains the integral of the intensity function, without any numerical approximations. Our model achieves competitive or superior performances compared to the previous state-of-the-art methods for both synthetic and real datasets.