Using Support Vector Machines to Predict the Performance of MLP Neural Networks

In this work, we investigated the use of support vector machines (SVM) to predict the performance of learning algorithms based on features of the learning problems, in a kind of meta-learning. Experiments were performed in a case study in which SVM regressors with different kernel functions were used to predict the performance of multi-layer perceptron (MLP) networks. The results obtained on a set of 50 learning problems revealed that the SVMs obtained better results in predicting the MLP performance,when compared to benchmark algorithms applied in previous work.