A byte-pair-encoding tokenizer that converts repeated temporal motifs into single tokens improves zero-shot forecasting accuracy and speed over sample-wise and patch-based methods.
An adaptive tokenization approach for time series We provide pseudocode for generating a vocabulary of motifs and utilizing the motifs to tokenize a time series
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Byte Pair Encoding for Efficient Time Series Forecasting
A byte-pair-encoding tokenizer that converts repeated temporal motifs into single tokens improves zero-shot forecasting accuracy and speed over sample-wise and patch-based methods.