“FIN Status Report 2017-2-20”版本间的差异

give a talk about understanding deep learning.
try RMSProp optimizer, weight decay and obtain smooth predicted sequences.
try industry stock and obtain highly correlated predicted sequence.
try to add index data and obtain more accurate predicted value.

study the stock selecting model.
study the architecture of the network.

Pick stable stocks and Leading stocks

Continue to pick Leading stocks and Make database smoothy to use

Implement:
regularization: L2-norm, dropout, batch normalization(including some bugs)
filters: MA, wavelet
Implement index data.
Improve strategy.

Improve strategy.