Research map / Learning theory and algorithms / Prediction and decision losses
Triadic-OCD: Asynchronous Online Change Detection with Provable Robustness, Optimality, and Convergence
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes Triadic-OCD, an asynchronous distributed online change detection method that handles uncertain system matrices via cutting-plane approximation and augmented-Lagrangian updates, with convergence analysis and IEEE 14-bus FDIA experiments.
Problem
Online change detection in dynamic systems (e.g., FDIA detection in smart grids) when the system matrix is only known within an uncertainty set and detection must be distributed across sub-regions without synchronous stragglers.
Approach
Combines a robust generalized CUSUM statistic with constraint-wise uncertainty sets, approximates the convex protection-function constraints with dynamically updated cutting planes, and solves the resulting nested problem with asynchronous distributed ADMM-style updates where the master updates after S of L workers respond.
Key results
- Triadic-OCD achieves smaller average detection delay than RGCUSUM, WCD, and Adaptic CUSUM across false alarm periods (500 Monte Carlo runs), per the reported plots.
- Triadic-OCD achieves higher successful detection rate than RGCUSUM and WCD over 10^3 random attack vectors.
- Asynchronous mode (S=10, tau=10) converges faster than its synchronous counterpart under log-normal worker delays LN(1,0.5).
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use Triadic-OCD instead of the compared detectors | FDIA detection on the IEEE-14 bus system · IEEE-14 bus FDIA detection | average detection delay at the same false alarm periods | Helped |
| Use Triadic-OCD instead of RGCUSUM and WCD | FDIA detection across randomly generated attack vectors · IEEE-14 bus FDIA detection | successful detection rate | Helped |
| Use asynchronous rather than synchronous variable updates | Triadic-OCD · Distributed optimization convergence | convergence speed | Helped |
Limitations
- Experiments limited to simulated IEEE 14-bus smart grid data with a single uncertain area
- Convergence analysis relies on strong bounded-variable and step-size assumptions
- Results reported only as plots without tabulated numbers in the extracted text
Experimental setup
Datasets: IEEE 14-bus system measurement data (simulated)
Benchmarks: IEEE 14-bus FDIA simulation
Metrics: average detection delay under false alarm period successful detection rate vs delay upper bound convergence speed
Baselines: Adaptive CUSUM RGCUSUM WCD synchronous Triadic-OCD
Builds on: CUSUM generalized likelihood ratio test robust optimization with uncertainty sets
Closest papers by meaning
- Optimal Online Change Detection via Random Fourier Features (NeurIPS 2025 · similarity 0.82)
- Online Adaptive Anomaly Thresholding with Confidence Sequences (ICML 2024 · similarity 0.81)
- Reducing sequential change detection to sequential estimation (ICML 2024 · similarity 0.81)
- Weight for Robustness: A Comprehensive Approach towards Optimal Fault-Tolerant Asynchronous ML (NeurIPS 2024 · similarity 0.81)
- Online Isolation Forest (ICML 2024 · similarity 0.79)
- Do We Need Asynchronous SGD? On the Near-Optimality of Synchronous Methods (NeurIPS 2026 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.