change-point-detection

Gaussian Process change-point detection identifies statistical property shifts in financial time-series to segment market regimes. It compares a Matérn 3/2 GP with a change-point kernel using marginal-likelihood to reveal distinct market states.

To detect market regime transitions, apply GP-CPD using lookback windows and min/max length constraints. This threshold-based segmentation workflow processes financial time-series to isolate meaningful market states and momentum crash detection points.

Yes, GP-CPD constructs context sets for few-shot learning by segmenting financial time-series into detected regimes. These identified market states support adaptive trading strategies and regime-based model training.

Yes, Gaussian Process change-point detection requires a Matérn 3/2 kernel. It compares this Matérn GP against a change-point kernel using marginal-likelihood evaluations to accurately identify regime shifts in financial time-series.

Marginal-likelihood comparisons for regime segmentation require careful threshold configuration and min/max length constraints. Without properly tuned lookback windows, detected change-points may produce unstable or overlapping market regime segments.