Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2501.14912 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912203493343232 |
|---|---|
| author | Ramirez, Juan Hounie, Ignacio Elenter, Juan Gallego-Posada, Jose Hashemizadeh, Meraj Ribeiro, Alejandro Lacoste-Julien, Simon |
| author_facet | Ramirez, Juan Hounie, Ignacio Elenter, Juan Gallego-Posada, Jose Hashemizadeh, Meraj Ribeiro, Alejandro Lacoste-Julien, Simon |
| contents | We introduce Feasible Learning (FL), a sample-centric learning paradigm where models are trained by solving a feasibility problem that bounds the loss for each training sample. In contrast to the ubiquitous Empirical Risk Minimization (ERM) framework, which optimizes for average performance, FL demands satisfactory performance on every individual data point. Since any model that meets the prescribed performance threshold is a valid FL solution, the choice of optimization algorithm and its dynamics play a crucial role in shaping the properties of the resulting solutions. In particular, we study a primal-dual approach which dynamically re-weights the importance of each sample during training. To address the challenge of setting a meaningful threshold in practice, we introduce a relaxation of FL that incorporates slack variables of minimal norm. Our empirical analysis, spanning image classification, age regression, and preference optimization in large language models, demonstrates that models trained via FL can learn from data while displaying improved tail behavior compared to ERM, with only a marginal impact on average performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_14912 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Feasible Learning Ramirez, Juan Hounie, Ignacio Elenter, Juan Gallego-Posada, Jose Hashemizadeh, Meraj Ribeiro, Alejandro Lacoste-Julien, Simon Machine Learning Artificial Intelligence We introduce Feasible Learning (FL), a sample-centric learning paradigm where models are trained by solving a feasibility problem that bounds the loss for each training sample. In contrast to the ubiquitous Empirical Risk Minimization (ERM) framework, which optimizes for average performance, FL demands satisfactory performance on every individual data point. Since any model that meets the prescribed performance threshold is a valid FL solution, the choice of optimization algorithm and its dynamics play a crucial role in shaping the properties of the resulting solutions. In particular, we study a primal-dual approach which dynamically re-weights the importance of each sample during training. To address the challenge of setting a meaningful threshold in practice, we introduce a relaxation of FL that incorporates slack variables of minimal norm. Our empirical analysis, spanning image classification, age regression, and preference optimization in large language models, demonstrates that models trained via FL can learn from data while displaying improved tail behavior compared to ERM, with only a marginal impact on average performance. |
| title | Feasible Learning |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2501.14912 |