I am working on a project using trainable WEKA to segment pathology in color retinal photographs. My goal is to train a classifier using multiple sample retinal images. I was wondering, is it necessary to reload multiple ARFF files to do this, or just the most recent?
For example, if for two separate images: I load my created classifier, train my classifier, and save two separate data files (as ARFF), for the next training image should I load the latest ARFF data file to train my classifier, or should I be loading both prior files before training? In other words, for training a classifier should I load data from previous images sequentially, or add all the cumulative data files?
Additionally, is it possible to adjust classifier settings at any point while training using multiple images, or are my setting choices locked-in the moment I create my classifier?