- The code is compatible for python 3.6 and tensorflow 1.1.
- The static RNN is deployed in the post LSTM_tsc and we adopt the dynamical RNN in tensorflow to achieve better computational speed.
- We further modify the batch process and add the GRU cells.
- For the ChlorineConcentration data set, applying the train-test (10%/90%) split discussed in this paper, it is easy to reach >75% test accuracy.
Credits for this project go to LSTM_tsc for providing a strong example and the UCR archive for the dataset.