Error initializing: covariance matrixes

[ellaksay]

Hello! We are processing a dataset with hidden state numerical values and additional features like distance and velocity. When we run the dataset with just the numerical values, the model works but when we run with the additional features, we get this error. How can we adjust our dataset to eliminate this error? We have no NaN values in our dataset, so unsure of the cause of this error.

[vidaurre]

It might be that there are extreme events in your data. For example, if you standardize each time series, are there time points which absolute value is several standard deviations above zero?

In general it's hard to know without seeing the data, but I would recommend some visual inspection. Maybe also run in subsets of the data to see when it happens.