Sequential Changepoint Detection via Backward Confidence Sequences

Authors: Shubhanshu Shekhar, Aaditya Ramdas

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We provide strong nonasymptotic guarantees on the frequency of false alarms and detection delay, and demonstrate numerical effectiveness on several problems. and Finally, in Section 5, we demonstrate the power and generality of our proposed scheme by instantiating it with five different confidence sequences. The general bound on the detection delay obtained in Theorem 13 easily translate into problem-specific upper bounds in all these cases, and we also empirically verify the theoretical predictions through some simple numerical simulations.
Researcher Affiliation Academia 1Department of Statistics and Data Science, Carnegie Mellon University, USA 2Machine Learning Department, Carnegie Mellon University, USA. Correspondence to: Shubhanshu Shekhar <shubhan2@andrew.cmu.edu>.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks (e.g., a figure or section labeled 'Algorithm' or 'Pseudocode').
Open Source Code Yes The code for reproducing the empirical results is available here.
Open Datasets No The paper mentions using 'Higgs, Banknote, and Occupancy' datasets and describes synthetic data generation, but does not provide concrete access information (specific links, DOIs, repositories, or formal citations with authors/year) for a publicly available or open dataset.
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment.
Experiment Setup Yes Empirical Verification. We now verify the theoretical claims of our proposed changepoint detection scheme using observations drawn from a unit-variance normal distribution with the pre-change mean θ0 = 0, and post-change mean θ1 = . In Figure 2, we consider the case of = 0.4 with the change occurring at T = 800. and We consider the following datasets for binary classification from the UCI Machine learning repository: Higgs, Banknote, and Occupancy. ... with a target ARL of 500 on all the three datasets.