LSTM gradient check fails

bayerj / arac

C++ library for neural networks.

Other

39 stars 57 forks source link

LSTM gradient check fails #5

Open bayerj opened 15 years ago

bayerj commented 15 years ago

And has to be corrected:

[ RUN      ] TestGradient.LstmLayer
New Parametrized instance.
Numeric/Analytical:-0 <-> 0
Numeric/Analytical:-0 <-> 0
Numeric/Analytical:-0 <-> 2.11224
Numeric/Analytical:-0 <-> 0
Expected: (0.001) > (gradient_check(*net_p, true)), actual: 0.001 vs 2.11224

However, the examples give the same solutions as PyBrain. Or?

bayerj commented 15 years ago

This is due to gradient_check not being able to handle sequential modules so far.