missing module DataSet about lstm-deepfm HOT 3 CLOSED

Thanks for your question, we'll reorganize the code and fix this in two days.

from lstm-deepfm.

1. We reconstructed the code base in order to make it more pythonic.
2. We discovered that, in LSTM-DeepFM/pretraining/selfsupervised_pretraining.py, row 94, the std as denominator may cause excessive losses, the std is dataset dependent.
3. We discovered that the scipy optimize can further optimize the performance of the model. FM is great useful if there is many discrete features.、

from scipy.optimize import minimize
from sklearn.metrics import mean_squared_error
def f(x):
   pred1 Deep  pred2 FM
   pred = x[0]*pred1 + x[1]*pred2
   score = mean_squared_error(pred, label)
    return score

weight_init = [0.5, 0.5]
result = minimize(f, weight_init , method="Nelder-Mead")
print(result.x)

4. We recommend the https://github.com/JustusvLiebig/Soft_Sensor_Experiments, since there is many implemented soft sensor models.

from lstm-deepfm.

Thanks, it works now.

from lstm-deepfm.