Author Verification: Exploring a Large set of Parameters using a Genetic Algorithm - Notebook for PAN at CLEF 2014
Erwan Moreau, Arun Jayapal, Carl M. Vogel · Arrow@dit (Dublin Institute of Technology) · 2014
Abstract In this paper we present the system we submitted to the PAN’14 com-petition for the author verification task. We consider the task as a supervised classification problem, where each case in a dataset is an instance. Our system works by applying the same combination of parameters to every case in a dataset. Thus, the training stage consists in finding an optimal combination of parameters which maximizes the performance on the training data using cross-validation. This is achieved using a simple genetic algorithm, since the space of all possible combinations is impractical. 1