Online learning partial least squares regression model for univariate response data

Partial least squares (PLS) analysis has attracted increasing attentions in image and video processing. Currently, most applications employ batch-form PLS methods, which require maintaining previous training data and re-training the model when new observations are available. In this work, we propose a novel approach that is able to update the PLS model in an online fashion. The proposed approach has the appealing property of constant computational complexity and const space complexity. Two extensions are proposed as well. First, we extend the method to be able to update the model when some training samples are removed. Second, we develop a weighted version, where different weights can be assigned to the data blocks when updating the model. Experiments on real image data confirmed the effectiveness of the proposed methods.