Learning adaptive kernels for model diagnosis
Bernardete Ribeiro · Hybrid Intelligent Systems · 2003
This paper looks into the tradeoff between model complexity and prediction accuracy using data examples from the benchmark problem of breast cancer. In particular, we take into account several crucial aspects in model construction using learning kernel classifiers. Given its importance, a more generalized form of the basis kernel function definition is then applied. Moreover, model selection is performed in terms of kernel machine hyperparameters, and results are evaluated in terms of the model development cost time.