numpy learning

One-hot labels preprocessing

• Weighting samples with confidence score

  1 2 3

  index = np.ragmax(predicted_result,axis=1) arg = np_utils.to_categorical(index,classes) weighted_one_hot=predicted_result*arg

• Selection && weighting samples with confidence score

  1	weighted_selected_one_hot=np.where(predicted_result>k,predicted_result,0)

• Extending vector into matrix

  1	xx = [np.full(classes,value) for value in x]