Table 7:

Validation Errors at the End of the Training Period for a Fully Connected One-Hidden-Layer Network Trained on MNIST, with Different γ Values.

γ value	Mean Validation Error (%)	Minimum Validation Error (%)	Maximum Validation Error (%)
0.2	2.75	2.64	2.85
0.5	2.51	2.43	2.60
0.7	2.38	2.24	2.47
0.8	2.41	2.32	2.47
0.9	2.26	2.21	2.31
1.0	2.45	2.38	2.53
1.1	2.46	2.37	2.63
1.2	2.35	2.26	2.48
1.3	2.43	2.28	2.54
1.5	83.16	83.16	83.16
1.8	90.09	90.09	90.09

Notes: For parameters that converged to low (<3%) validation errors, four trials were run, and the mean, minimum, and maximum errors over the trials have been reported. In all runs, β = 1.