I need to check for stationarity of the time series. How can this be done using tensorflow?
You can reference this tutorial on predicting time series data: Time series forecasting | TensorFlow Core
It does not require to check for stationarity before training the model. However, in some cases the model will benefit if you generate rolling and cumulative features based on the original time series values.
If you asked about the functions similar to those that are available in statsmodels, I did not come across a ready solutions. TF is a huge library. Probably you can do something with tf.math.segment_mean function or other math functions.
Tensorflow Probabilty has also a statistical component.
TensorFlow Probability is a library for probabilistic reasoning and statistical analysis.:
You can find also two Timeseries tutorials at:
Yes, I’m just studying now tensorflow_probability, I want to find Dickie Fuller’s tests or KPSS