In-context Time Series Predictor
This addresses overfitting and inefficiency in time series forecasting models, offering a parameter-efficient solution for practitioners in fields like finance or weather prediction, though it is incremental as it builds on existing in-context learning concepts.
The paper tackled time series forecasting by reformulating tasks as input tokens using (lookback, future) pairs to better align with in-context learning, achieving consistently better performance across full-data, few-shot, and zero-shot settings compared to previous architectures.
Recent Transformer-based large language models (LLMs) demonstrate in-context learning ability to perform various functions based solely on the provided context, without updating model parameters. To fully utilize the in-context capabilities in time series forecasting (TSF) problems, unlike previous Transformer-based or LLM-based time series forecasting methods, we reformulate "time series forecasting tasks" as input tokens by constructing a series of (lookback, future) pairs within the tokens. This method aligns more closely with the inherent in-context mechanisms, and is more parameter-efficient without the need of using pre-trained LLM parameters. Furthermore, it addresses issues such as overfitting in existing Transformer-based TSF models, consistently achieving better performance across full-data, few-shot, and zero-shot settings compared to previous architectures.