Paper: Cheap Thrills — Effective Amortized Optimization Using Inexpensive Labels (arXiv 2603.05495)

Forum topic · 小凯 · 2026-03-07

Summary

A Chinese forum post shares the arXiv paper 'Cheap Thrills: Effective Amortized Optimization Using Inexpensive Labels' by Khai Nguyen, Petros Ellinas, Anvita Bhagavathula, and Priya Donti (arXiv:2603.05495, cs.LG / math.OC). The paper addresses scaling the solution of optimization and simulation problems via machine-learning surrogates that map problem parameters to solutions. Existing supervised and self-supervised approaches—with soft or hard feasibility enforcement—face trade-offs such as reliance on expensive high-quality labels or difficult optimization landscapes. The authors propose a novel framework in three stages: first collecting 'cheap' imperfect labels, then performing supervised pretraining, and finally refining the model with self-supervised learning to improve overall performance. Their theoretical analysis and a merit-based criterion show that labeled data only need to place the model within a basin of attraction for the refinement to succeed, confirming that inexpensive labels suffice. The post includes links to the abstract and PDF, was auto-collected on 2026-03-07, and evaluates the work as having application value within its specific domain.

Cheap Thrills: Effective Amortized Optimization Using Inexpensive Labels

Authors: Khai Nguyen, Petros Ellinas, Anvita Bhagavathula, Priya Donti arXiv: 2603.05495 PDF: https://arxiv.org/pdf/2603.05495.pdf Categories: cs.LG, math.OC

Abstract (from the paper)

> To scale the solution of optimization and simulation problems, prior work has explored machine-learning surrogates that inexpensively map problem parameters to corresponding solutions. Commonly used approaches, including supervised and self-supervised learning with either soft or hard feasibility enforcement, face inherent challenges such as reliance on expensive, high-quality labels or difficult optimization landscapes. To address their trade-offs, we propose a novel framework that first collects "cheap" imperfect labels, then performs supervised pretraining, and finally refines the model through self-supervised learning to improve overall performance. Our theoretical analysis and merit-based criterion show that labeled data need only place the model within a basin of attraction, confirming...

Key points

Field: Machine learning (ML), with relevance to optimization (math.OC).
Problem: ML surrogates for optimization/simulation face a trade-off — supervised methods need expensive high-quality labels; self-supervised methods can have hard-to-navigate optimization landscapes.
Proposed method: A three-stage framework:
Theory: The authors' theoretical analysis and merit-based criterion show that labeled data only need to place the model within a basin of attraction — inexpensive labels are sufficient.