this version seems to work

dicarlolab / archconvnets

Architecturally optimized neural networks trained with regularized backpropagation

5 stars 5 forks source link

Closed ardila closed 10 years ago

ardila commented 10 years ago

This branch definitely works: I have gotten a square error of 2 (summed over all IT units) on training data. @yamins81 I am merging.