Most powerful test sequences with early stopping options

ObjectiveIn patients with chronic coronary or peripheral artery disease enrolled in the Cardiovascular Outcomes for People Using Anticoagulation Strategies trial, randomised antithrombotic treatments were stopped after a median follow-up of 23 months because of benefits of the combination of rivaroxaban 2.5 mg two times per day and aspirin 100 mg once daily compared with aspirin 100 mg once daily. We assessed the effect of switching to non-study aspirin at the time of early stopping.MethodsIncident composite of myocardial infarction, stroke or cardiovascular death was estimated per 100 person-years (py) during randomised treatment (n=18 278) and after study treatment discontinuation to non-study aspirin (n=14 068).ResultsDuring randomised treatment, the combination compared with aspirin reduced the composite (2.2 vs 2.9/100 py, HR: 0.76, 95% CI 0.66 to 0.86), stroke (0.5 vs 0.8/100 py, HR: 0.58, 95% CI 0.44 to 0.76) and cardiovascular death (0.9 vs 1.2/100 py, HR: 0.78, 95% CI 0.64 to 0.96). During 1.02 years after early stopping, participants originally randomised to the combination compared with those randomised to aspirin had similar rates of the composite (2.1 vs 2.0/100 py, HR: 1.08, 95% CI 0.84 to 1.39) and cardiovascular death (1.0 vs 0.8/100 py, HR: 1.26, 95% CI 0.85 to 1.86) but higher stroke rate (0.7 vs 0.4/100 py, HR: 1.74, 95% CI 1.05 to 2.87) including a significant increase in ischaemic stroke during the first 6 months after switching to non-study aspirin.ConclusionDiscontinuing study rivaroxaban and aspirin to non-study aspirin was associated with the loss of cardiovascular benefits and a stroke excess.Trial registration numberNCT01776424.

Download Full-text

Quit When You Can: Efficient Evaluation of Ensembles by Optimized Ordering

ACM Journal on Emerging Technologies in Computing Systems ◽

10.1145/3451209 ◽

2021 ◽

Vol 17 (4) ◽

pp. 1-20

Author(s):

Serena Wang ◽

Maya Gupta ◽

Seungil You

Keyword(s):

Optimization Problem ◽

Optimal Solution ◽

Combinatorial Optimization Problem ◽

Joint Optimization ◽

Early Stopping ◽

Time Approximation ◽

Approximation Bound ◽

Evaluation Time ◽

Speed Up ◽

Real World Problems

Given a classifier ensemble and a dataset, many examples may be confidently and accurately classified after only a subset of the base models in the ensemble is evaluated. Dynamically deciding to classify early can reduce both mean latency and CPU without harming the accuracy of the original ensemble. To achieve such gains, we propose jointly optimizing the evaluation order of the base models and early-stopping thresholds. Our proposed objective is a combinatorial optimization problem, but we provide a greedy algorithm that achieves a 4-approximation of the optimal solution under certain assumptions, which is also the best achievable polynomial-time approximation bound. Experiments on benchmark and real-world problems show that the proposed Quit When You Can (QWYC) algorithm can speed up average evaluation time by 1.8–2.7 times on even jointly trained ensembles, which are more difficult to speed up than independently or sequentially trained ensembles. QWYC’s joint optimization of ordering and thresholds also performed better in experiments than previous fixed orderings, including gradient boosted trees’ ordering.

Download Full-text

Event-based Neural Network for ECG Classification with Delta Encoding and Early Stopping

2020 6th International Conference on Event-Based Control, Communication, and Signal Processing (EBCCSP) ◽

10.1109/ebccsp51266.2020.9291357 ◽

2020 ◽

Author(s):

Matthias Jobst ◽

Chen Liu ◽

Johannes Partzsch ◽

Yexin Yan ◽

David Kappel ◽

...

Keyword(s):

Neural Network ◽

Early Stopping ◽

Event Based

Download Full-text

An optimal early stopping uniform consensus protocol in synchronous distributed systems with orderly crash features

23rd International Conference on Distributed Computing Systems Workshops, 2003. Proceedings. ◽

10.1109/icdcsw.2003.1203536 ◽

2004 ◽

Author(s):

Xianbing Wang ◽

Jiannong Cao

Keyword(s):

Distributed Systems ◽

Early Stopping ◽

Consensus Protocol

Download Full-text

Power comparison of Cochran-Armitage trend test against allelic and genotypic tests in large-scale case-control genetic association studies

Statistical Methods in Medical Research ◽

10.1177/0962280216683979 ◽

2016 ◽

Vol 27 (9) ◽

pp. 2657-2673 ◽

Cited By ~ 2

Author(s):

Mathieu Emily

Keyword(s):

Large Scale ◽

Association Studies ◽

Disease Model ◽

Trend Test ◽

Genome Wide Association Studies ◽

Power Functions ◽

Power Comparison ◽

Powerful Test ◽

Armitage Trend Test ◽

Mode Of Inheritance

The Cochran-Armitage trend test (CA) has become a standard procedure for association testing in large-scale genome-wide association studies (GWAS). However, when the disease model is unknown, there is no consensus on the most powerful test to be used between CA, allelic, and genotypic tests. In this article, we tackle the question of whether CA is best suited to single-locus scanning in GWAS and propose a power comparison of CA against allelic and genotypic tests. Our approach relies on the evaluation of the Taylor decompositions of non-centrality parameters, thus allowing an analytical comparison of the power functions of the tests. Compared to simulation-based comparison, our approach offers the advantage of simultaneously accounting for the multidimensionality of the set of features involved in power functions. Although power for CA depends on the sample size, the case-to-control ratio and the minor allelic frequency (MAF), our results first show that it is largely influenced by the mode of inheritance and a deviation from Hardy–Weinberg Equilibrium (HWE). Furthermore, when compared to other tests, CA is shown to be the most powerful test under a multiplicative disease model or when the single-nucleotide polymorphism largely deviates from HWE. In all other situations, CA lacks in power and differences can be substantial, especially for the recessive mode of inheritance. Finally, our results are illustrated by the comparison of the performances of the statistics in two genome scans.

Download Full-text