To model temporal dependencies, a sliding window of width w is used to concatenate consecutive observations into a high-dimensional vector, forming a subsequence (Equation 2). Here, w denotes the ...