Iteration over event space in time-to-first-spike spiking neural networks for Twitter bot classification

TL;DR

This paper introduces an extended time-to-first-spike SNN framework capable of processing temporally evolving information, demonstrated on Twitter bot detection with complex spike train data across multiple timescales.

Contribution

It presents a novel extension of SNN models to handle dynamic temporal data and develops training rules for end-to-end backpropagation in this context.

Findings

Effective processing of time-varying spike train data

Model handles events across five orders of magnitude in timescales

Analysis of parameter effects on performance and stability

Abstract

This study proposes a framework that extends existing time-coding time-to-first-spike spiking neural network (SNN) models to allow processing information changing over time. We explain spike propagation through a model with multiple input and output spikes at each neuron, as well as design training rules for end-to-end backpropagation. This strategy enables us to process information changing over time. The model is trained and evaluated on a Twitter bot detection task where the time of events (tweets and retweets) is the primary carrier of information. This task was chosen to evaluate how the proposed SNN deals with spike train data composed of hundreds of events occurring at timescales differing by almost five orders of magnitude. The impact of various parameters on model properties, performance and training-time stability is analyzed.