search

PeterChe1990 / GRU-D

GRU-D, a GRU-based model with trainable decays for multivariate time series classification with missing values/irregular samplings
MIT License
118 stars 37 forks source link

Access the imputed dataset #7

Closed noyomer closed 1 year ago

noyomer commented 1 year ago

Hi, Thanks for this great work! I successfully ran the GRUD on samples (using run.ipynb). I am interested in the imputation output (X' with the imputed values), but it seems that only the prediction performance results are stored. Could you please guide me on how to access the imputed dataset?

Thanks!

PeterChe1990 commented 1 year ago

Hi moyomer,

As mentioned in the paper, The proposed method is not explicitly designed for filling in the missing values in the data, and can not be directly used in unsupervised settings without prediction labels. (https://www.nature.com/articles/s41598-018-24271-9#Sec22)

In the case that the imputed data is indeed needed, you may modify the model and take the output from the last GRU-D layer as the imputation. https://github.com/PeterChe1990/GRU-D/blob/9e1274a1ad67135137f53159eafc92c7278a931a/models.py#L56 (i.e., before the first MLP layer)

noyomer commented 1 year ago

Thank you!