noise images not predicted in CLASSIFICATION_STABILITY method

[alixunxing]

In active_training.py
line 167 ~ 169:
noise = np.round( np.random.normal(mu, sigma, X_Pool_subset.shape) ) aug_X_Pool_subset = X_Pool_subset + noise np.clip(aug_X_Pool_subset, 0, 255)

but in line 172:
curr_sample = [MC_output([X_Pool_subset, learning_phase])[0]]
the raw data X_Pool_subset is predicted.

I revised line 172 to:
curr_sample = [MC_output([aug_X_Pool_subset, learning_phase])[0]]
and run training, but the result is bad......

[CtfChan]

Hi @PoisonBOx thanks for pointing this out. I'll fix this typo :P