Biomedical Image Classification with Multi Response Linear Regression (MLR) as Meta-Learner Combiner and Its Effectiveness on Small to Large Data Sets
Md Mahmudur Rahman, Prabir Bhattacharya · 2016
This paper presents a multi-modal biomedical image classification approach by using a multi-response linear regression (MLR)-based meta-learner as combiner and measure its effectiveness from small to large image collections. The MLR has been proposed as a trainable combiner for fusing class probability outputs of several base-level SVM classifiers on multiple complimentary visual and text features as inputs. The advantage of using MLR here over other generalizers is its interpretability as the weights generated by it indicate the different contributions that each features makes for class prediction. The usability of MLR as a meta-learner combiner is evaluated by performing experimental analysis with data sets of varying sizes (small and large samples) and its effectiveness is established based on the significant performance improvement in result analysis.