Classification Using Generalized Partial Least Squares

摘要: Advances in computational biology have made simultaneous monitoring of thousands features possible. The high throughput technologies not only bring about a much richer information context which to study various aspects gene function, but they also present the challenge analyzing data with large number covariates and few samples. As an integral part machine learning, classification samples into two or more categories is almost always interest scientists. We address question this setting by extending partial least squares (PLS), popular dimension reduction tool chemometrics, generalized linear regression, based on previous approach, iteratively reweighted squares, that is, IRWPLS. compare our results two-stage PLS other classifiers. show phrasing problem model applying Firth's procedure avoid (quasi)separation, we often get lower classificatio...