Adaptively Truncated Signature-based Logistic Regression for Semi-parametric Functional Classification
This research introduces Path Signatures Logistic Regression (PSLR) to classify complex functional data using scalar covariates. It combines a semi-parametric additive structure with a data-driven method to automatically select the optimal signature truncation order. This allows the model to adapt its complexity based on the data, providing robust classification while maintaining interpretable effects for each input variable.