Introducing Multivariate Anomaly Detection

Great article. I just had a couple of questions?

1) I see a preprocessing block. What kind of data preprocessing does it support? Also, does it support preprocessing during training as well as inference?

2) After training a model what metric does the service provide for verifying the accuracy?

3) Do you have limitations on the number of sensors or timestamps supported? Do you have any metric on the latency of training or inference?