DICE is now mainly applicable to two-dimensional tabular data, e.g., N*M.
How can I apply DICE to a three-dimensional time-series dataset (N*T*M)?
This can be used with LSTM or Transformers that require three-dimensional inputs.