vanderschaarlab / clairvoyance Goto Github PK

Not directly a clairvoyance problems but some missingpy 0.2.0 components try to import sklearn.neighbors.base which isn’t right anymore as that’s now been renamed sklearn.neighbors._base since sklearn 0.22.1 I believe. Have reported issue to missingpy but work around is to modify missingpy components directly.

Hi,
Just wanted to ask if there are any plans to include other models into the package that your lab has developed. Specifically Dynamic-Deephit and DCN-PD would be great to have.

In CRN_Base.py,

from tensorflow.contrib.rnn import LSTMCell, DropoutWrapper
does not work because it has been moved.

Instead call (at least for now)

from tensorflow.compat.v1.nn.rnn_cell import LSTMCell

The gain_imputation.py code tries to import deprecated tensorflow.saved_model.tag_constants.

Fortunately the tags don't seem to be required as far as I can see so

tf.saved_model.loader.load(sess, [tag_constants.SERVING], self.save_file_directory)

can be replaced by

tf.saved_model.loader.load(sess, self.save_file_directory)

Would it be possible to share the scripts to get the mimic dataset needed for the tutorials about prediction and active sensing. It uses to "normal" mimic dataset, but only for the mimim_antibiotics dataset the data extraction script is given.

It's a wonderful package and very well written and documented ;) Congrats to that.

Best wishes, Elly

Thanks so much for your work on this!

I'm trying to run the tutorial here, but the link to https://github.com/jsyoon0823/time-series-automl.git is dead. Any updates on that? :-)

Thanks!

In treatments/RMSN/libs/net_helpers.py the line

def get_optimization_graph(loss, learning_rate, max_global_norm, global_step, optimisation_function=tf.train.AdamOptimizer):

tf.train.AdamOptimizer is deprecated: should now be tf.train.Adam()

i.e.

def get_optimization_graph(loss, learning_rate, max_global_norm, global_step, optimisation_function=tf.optimizers.Adam()):

Small issue but it would be good to have an empty '~/clairvoyance/datasets/data/mimic_antibiotics' directory set up in the repo as a placeholder (extract_antibiotics_dataset.py script fails otherwise).

:-)

In treatments/RMSN/libs/model_rnn.py,

tf.contrib.rnn.RNNCell has been moved to tf.compat.v1.nn.rnn_cell.RNNCell (couple of defs:)