Key points are not available for this paper at this time.
Most machine learning researchers perform quantitative experiments to estimate generalization error and compare the performance of different algorithms (in particular, their proposed algorithm). In...
BengioYoshua et al. (2004) studied this question.