Performance on regression tasks as shown by MLPs trained on some exemplar UCI datasets.

After reporting relevant information for each dataset (i.e., number of instances, number of features and the output signal selected for regression), the table shows the performance measured in terms of R2 and δ. The difference (in absolute value) between R2 and δ has also been reported. Experiments have been performed by repeating 10 training and test runs (each with 70% and 30% of available examples), followed by averaging. Each test has been performed with random sampling without replacement. Problems are also labeled with one or two white circles (standing for low and medium difficulty, respectively).