Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2606.01557 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913177941311488 |
|---|---|
| author | Boero, Ignacio Hounie, Ignacio Chamon, Luiz Ribeiro, Alejandro |
| author_facet | Boero, Ignacio Hounie, Ignacio Chamon, Luiz Ribeiro, Alejandro |
| contents | Everywhere learning is a new paradigm whereby Artificial Intelligence (AI) systems are trained to satisfy loss constraints with probability one over the data distribution. This is in contrast to the standard paradigm of training AI systems to minimize average losses. We develop an approximate duality theory to substantiate a generalization analysis that establishes the proximity between solutions of empirical and statistical everywhere learning problems. Our results show that dual variables reweigh the data distribution towards points in which loss constraints are more difficult to satisfy and that generalization is controlled by the mismatch between the concentration of mass of the data distribution and the concentration of mass on points where constraints are more difficult to satisfy. We further show that we can control generalization with a sparse L1 penalty on constraint relaxations. We illustrate the merits of everywhere learning with an experiment in agentic classification for language model tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2606_01557 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Everywhere Learning: Artificial Intelligence with Pointwise Constraints Boero, Ignacio Hounie, Ignacio Chamon, Luiz Ribeiro, Alejandro Machine Learning Signal Processing Everywhere learning is a new paradigm whereby Artificial Intelligence (AI) systems are trained to satisfy loss constraints with probability one over the data distribution. This is in contrast to the standard paradigm of training AI systems to minimize average losses. We develop an approximate duality theory to substantiate a generalization analysis that establishes the proximity between solutions of empirical and statistical everywhere learning problems. Our results show that dual variables reweigh the data distribution towards points in which loss constraints are more difficult to satisfy and that generalization is controlled by the mismatch between the concentration of mass of the data distribution and the concentration of mass on points where constraints are more difficult to satisfy. We further show that we can control generalization with a sparse L1 penalty on constraint relaxations. We illustrate the merits of everywhere learning with an experiment in agentic classification for language model tasks. |
| title | Everywhere Learning: Artificial Intelligence with Pointwise Constraints |
| topic | Machine Learning Signal Processing |
| url | https://arxiv.org/abs/2606.01557 |