An Iterative Dynamic Ensemble Weighting Approach for Deep Learning Applications

For deep learning applications, large numbers of samples are essential. If this condition is not met, effective features cannot be generated and overfitting occurs especially for the small datasets such as in medical applications. To address this issue, we propose a new dynamic ensemble merging algorithm that iteratively adjusts the weights of a convolutional neural network (CNN) ensemble's elements in an online manner. For given test instance, the proposed algorithm(1), initially assigns equal weights to each of the classifiers and increases the weights of best k ones along iterations. Experiments that we conduct on a small deep learning dataset lead to promising ensemble results compared to its counterparts.