Abstract
In this work, an improvement is made for incremental principal component analysis (IPCA) in which the addition of a new eigen-axis is determined based on the accumulation ratio for a currently constructed feature space. In incremental learning situations, the accumulation ratio must be modified every time a new sample is given. Therefore, in order to use this ratio as a criterion for the addition of a new feature in IPCA, we need to develop a one-pass update algorithm for the accumulation ratio. To obtain its approximation without using all the past samples, we present an improved algorithm of IPCA. To see if correct feature construction is carried out by the new IPCA algorithm, the recognition performance is evaluated for some standard datasets when k-NN classifier is adopted as a classifier.
