fastcan: A fast canonical-correlation-based searching algorithm

Searching techniques play a critical role in machine learning and data science, particularly for tasks such as feature selection and data pruning. This paper introduces fastcan , a highly efficient Python implementation of the canonical-correlation-based searching algorithm, and its impact on machine learning research. Designed for both speed and flexibility, fastcan achieves high computational efficiency through novel mathematical derivations and Cython-based optimisations. It supports multi-output scenarios and incorporates an automatic mechanism to exclude similar or redundant data, enhancing the quality of selected features. Moreover, fastcan includes a dedicated submodule, narx , which facilitates time-series regression via rapid and effective regressor selection. The speed, flexibility and specialised characteristics of fastcan make it a valuable tool for benchmarking feature selection techniques, data pruning methods, and time-series regression models.