issues
search
accosmin
/
nano
C++ library [machine learning & numerical optimization] - superseeded by libnano
MIT License
1
stars
0
forks
source link
Update the training scripts
#58
Closed
accosmin
closed
8 years ago
accosmin
commented
8 years ago
Just for MNIST:
train some MLP and convolution models
vary the training method (batch CGD/LBFGS, minibatch CGD/LBFGS, stochastic)
vary the regularization method (AVG, L2, VAR)
simplify/cleanup/document the training script
construct nice plots to compare the performance of each method
Just for MNIST: