I need to predict technical aggregate condition using vibration monitoring data. We consider this data to be nonstationary i.e. distribution parameters and descriptive statistics are not constant. I found that one of the best algorithms for such tasks in Learn++.NSE and we us it with MLP as a base classifier.

As I know, it's necessary no normalize data for operations with ANN. We decided to normalize using mean, stdev and sigmoidal function. We train networks of ensemble with sets with different values distribution parameters.

So, my questions are the following

  1. How to normalize new trainig set during previous networks evaluation? Problem is in description statistics change
  2. How to normalize input data while ensemble usage? Current statistics differ from the previous ones.

Batch normalization is critical technique for fast learning speed and generalization [8]. In this paper, batch temporal normalization layer is proposed for stationarity of input time series.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.