Maria-Eleni Sfyraki, Jun-Kun Wang · 2026 · arXiv (Cornell University)
DOI: 10.48550/arxiv.2610.26790 · cited by 0 works (OpenAlex) · open-access PDF
O treinamento de um modelo de aprendizado de máquina (ML) é um processo de otimização que normalmente requer um conjunto de exemplos com rótulos de verdade básica antecipadamente. Obter essas verdades básicas geralmente requer esforço humano laborioso e pode ser dispendioso. A escalabilidade da rotulagem humana torna-se uma preocupação quando há um grande número de pontos de dados que precisam de ser rotulados. Para resolver este problema prático, consideramos uma configuração de otimização onde um subconjunto de amostras tem rótulos de verdade básica, enquanto os pontos de dados restantes foram rotulados por um preditor de ML, o que poderia potencialmente atribuir rótulos incorretos a esses pontos. Proponemos métodos de otimização estocástica de primeira ordem e de ordem zero para o treinamento de conjuntos de dados consist
Machine translation (Meta M2M100); the English original follows.
Training a machine learning (ML) model is an optimization process that typically requires a set of examples with ground-truth labels beforehand. Obtaining these ground truths usually requires laborious human effort and can be costly. The scalability of human labeling becomes a concern when there are a large number of data points that need to be labeled. To address this practical issue, we consider an optimization setting where a subset of samples has ground-truth labels, while the remaining data points were labeled by an ML predictor, which could potentially assign incorrect labels to those points. We propose both first-order and zeroth-order stochastic optimization methods for training on datasets consisting of a mix of true-labeled and ML-labeled samples. We further conduct comprehensive empirical studies from training a linear regression model to fine-tuning a large language model (LLM) to showcase the effectiveness of the proposed algorithms for training with ML-labeled data.
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
Topics: Computer science, Machine learning, Training set, Training (meteorology), Process (computing), Set (abstract data type)
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.26790, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange