Zhiyong Zhang 15-05-24

From cslt Wiki
Jump to: navigation, search

Weekly Report:

  1, Finished hidden-layer dark-knowlede experiments on aurora4;
  2, Finished some pre-trianing (RBM and layer-by-layer-standard DNN)experiments on aurora4;
  3, Do some experiments on car-number recognition.

Target of next week:

  1, Continue the car-number recognition experiments; 
  2, Continue Hessian-like Monmentum training strategy including Classic-Momentum(CG) and Nesterov's Accelerated Gradient(NAG).(HOLD)