Collected sources and patterns will appear here. Add from search or the patterns library.
TimeSeriesTensor[Batch, SeqLen, Channels] -> PatchedTensor[Batch, NumPatches, PatchLen, Channels]
Segment a continuous time-series sequence into overlapping or non-overlapping windows (patches) to serve as discrete tokens for sequence models.
Problem it solves
Standard time-series points lack semantic density; patching reduces sequence length and captures local temporal context.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.