PPI-SVRG: Unifying Prediction-Powered Inference and Variance Reduction for Semi-Supervised Optimization

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Ao, Ruicheng, Chen, Hongyu, Liu, Haoyang, Simchi-Levi, David, Sun, Will Wei
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908796429795328
author Ao, Ruicheng
Chen, Hongyu
Liu, Haoyang
Simchi-Levi, David
Sun, Will Wei
author_facet Ao, Ruicheng
Chen, Hongyu
Liu, Haoyang
Simchi-Levi, David
Sun, Will Wei
contents We study semi-supervised stochastic optimization when labeled data is scarce but predictions from pre-trained models are available. PPI and SVRG both reduce variance through control variates -- PPI uses predictions, SVRG uses reference gradients. We show they are mathematically equivalent and develop PPI-SVRG, which combines both. Our convergence bound decomposes into the standard SVRG rate plus an error floor from prediction uncertainty. The rate depends only on loss geometry; predictions affect only the neighborhood size. When predictions are perfect, we recover SVRG exactly. When predictions degrade, convergence remains stable but reaches a larger neighborhood. Experiments confirm the theory: PPI-SVRG reduces MSE by 43--52\% under label scarcity on mean estimation benchmarks and improves test accuracy by 2.7--2.9 percentage points on MNIST with only 10\% labeled data.
format Preprint
id arxiv_https___arxiv_org_abs_2601_21470
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle PPI-SVRG: Unifying Prediction-Powered Inference and Variance Reduction for Semi-Supervised Optimization
Ao, Ruicheng
Chen, Hongyu
Liu, Haoyang
Simchi-Levi, David
Sun, Will Wei
Machine Learning
Econometrics
Optimization and Control
We study semi-supervised stochastic optimization when labeled data is scarce but predictions from pre-trained models are available. PPI and SVRG both reduce variance through control variates -- PPI uses predictions, SVRG uses reference gradients. We show they are mathematically equivalent and develop PPI-SVRG, which combines both. Our convergence bound decomposes into the standard SVRG rate plus an error floor from prediction uncertainty. The rate depends only on loss geometry; predictions affect only the neighborhood size. When predictions are perfect, we recover SVRG exactly. When predictions degrade, convergence remains stable but reaches a larger neighborhood. Experiments confirm the theory: PPI-SVRG reduces MSE by 43--52\% under label scarcity on mean estimation benchmarks and improves test accuracy by 2.7--2.9 percentage points on MNIST with only 10\% labeled data.
title PPI-SVRG: Unifying Prediction-Powered Inference and Variance Reduction for Semi-Supervised Optimization
topic Machine Learning
Econometrics
Optimization and Control
url https://arxiv.org/abs/2601.21470