Skip to Main Content
Table 2:
Classification Error Rates on MNIST Test Set When Only a Small Fraction of Labeled Data Is Available But the Remainder of the 60,000 Training Cases Is Available Unlabeled.
Two-LayerNonlinearLinearStack of
DBMNCANCAAutoencodersKNN
1% (600) 4.82% 8.81% 19.37% 8.13% 13.74% 
5% (3000) 2.72% 3.24% 7.23% 3.54% 7.19% 
10% (6000) 2.46% 2.58% 4.89% 2.98% 5.87% 
100% (60000) 0.95% 1.00% 2.45% 1.40% 3.09% 
Two-LayerNonlinearLinearStack of
DBMNCANCAAutoencodersKNN
1% (600) 4.82% 8.81% 19.37% 8.13% 13.74% 
5% (3000) 2.72% 3.24% 7.23% 3.54% 7.19% 
10% (6000) 2.46% 2.58% 4.89% 2.98% 5.87% 
100% (60000) 0.95% 1.00% 2.45% 1.40% 3.09% 
Close Modal

or Create an Account

Close Modal
Close Modal