tensor_utils

Top-level package imports for clean user-facing APIs.

tensor_utils.add_past_windows(A, lb, strict_lookback, axis=0)[source]

Build a rolling window of past values along one axis of a tensor.

Parameters:
  • A (ndarray) – An n-dimensional tensor.

  • lb (int) – The rolling window size.

  • strict_lookback (bool) – If True, exclude the value at timestamp T from the rolling window constructed for time T.

  • axis (int) – The axis along which to compute the rolling window.

Returns:

An (n+1)-dimensional tensor, A_dimsxW, with the last dimension being the window’s dimension.

Return type:

ndarray

tensor_utils.add_future_windows(A, lb, strict_lookahead, axis=0)[source]

Build a rolling window of future values along one axis of a tensor.

Parameters:
  • A (ndarray) – a n-dimensional tensor.

  • lb (int) – the rolling windows size.

  • strict_lookahead (bool) – If True, exclude the value at timestamp T from the rolling window constructed for time T.

  • axis (int) – the axis along which to compute the rolling window.

Returns:

An (n+1)-dimensional tensor, A_dimsxW, with the last dimension being the window’s dimension.

Return type:

ndarray