Sequential change detection via backward confidence sequences
This work provides a general solution for sequential changepoint detection, benefiting researchers and practitioners in fields like statistics and machine learning, though it is incremental as it builds on existing confidence sequence methods.
The paper tackles the problem of sequential changepoint detection by introducing a reduction from sequential estimation, using confidence sequences to detect changes in parameters or functionals of a distribution. The result includes strong nonasymptotic guarantees on false alarms and detection delay, with demonstrated numerical effectiveness across several problems.
We present a simple reduction from sequential estimation to sequential changepoint detection (SCD). In short, suppose we are interested in detecting changepoints in some parameter or functional $θ$ of the underlying distribution. We demonstrate that if we can construct a confidence sequence (CS) for $θ$, then we can also successfully perform SCD for $θ$. This is accomplished by checking if two CSs -- one forwards and the other backwards -- ever fail to intersect. Since the literature on CSs has been rapidly evolving recently, the reduction provided in this paper immediately solves several old and new change detection problems. Further, our "backward CS", constructed by reversing time, is new and potentially of independent interest. We provide strong nonasymptotic guarantees on the frequency of false alarms and detection delay, and demonstrate numerical effectiveness on several problems.