Sharp Regret Bounds and a Task-Covariance Correction for Spectral Representation Learning
Abstract
Spectral features can remain optimal under strongly uneven task preferences when they retain the directions most useful to the tasks. In a local-task model, expected probing gain depends on the task prior only through its covariance Λ: with B recording dependence between views, k spectral features span the leading eigenspace of BB⊤, while task-optimal features span that of BΛB⊤. We derive alignment-dependent regret bounds, matching worst-case lower bounds for a flat leading spectrum, and bounds using the leading 2k directions with a spectral-tail term; a spectral gap makes the bound quadratic in small anisotropy. We also bound the imbalance from random preference patterns and from averaging independent tasks with an isotropic population covariance. With known Λ, changing one term of the spectral contrastive loss selects task-optimal features; with a labelled task bank, we propose a diagnostic and test correction without retraining. Correction reduces synthetic held-out regret from $0.858$ to $0.003$ with $2000$ tasks when less task-relevant directions dominate, and on CIFAR-100 lowers empirical regret by $0.031−−0.053$ on held-out fine-label tasks using the same images. With a small task bank, however, correction can hurt.
Related
- Regret Equals Covariance: A Closed-Form Characterization for Stochastic Optimization
- Dual-domain self-supervised feature alignment via spectral–spatial representation learning for deepfake anomaly detection
- Nonparametric covariance estimation with bias correction for spatial lattice processes
- Hybrid Clifford Codes via Operator Algebra Quantum Error Correction and Projective Representation Theory
- Hamiltonian Lift of Bures--Wasserstein Covariance Dynamics with a Spectral Floor
BibTeX
@misc{tang2026sharp,
title = {{Sharp Regret Bounds and a Task-Covariance Correction for Spectral Representation Learning}},
author = {Dier Tang and Jing Yee Tan and Guangyue Han},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.15825},
eprint = {2609.15825},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.15825},
}
RIS
TY - UNPB
TI - Sharp Regret Bounds and a Task-Covariance Correction for Spectral Representation Learning
AU - Dier Tang
AU - Jing Yee Tan
AU - Guangyue Han
PY - 2026
DA - 2026/09/26
PB - arXiv
DO - 10.48550/arxiv.2609.15825
UR - https://arxiv.org/abs/2609.15825
AB - Spectral features can remain optimal under strongly uneven task preferences when they retain the directions most useful to the tasks. In a local-task model, expected probing gain depends on the task prior only through its covariance Λ: with B recording dependence between views, k spectral features span the leading eigenspace of BB^⊤, while task-optimal features span that of BΛB^⊤. We derive alignment-dependent regret bounds, matching worst-case lower bounds for a flat leading spectrum, and bounds using the leading 2k directions with a spectral-tail term; a spectral gap makes the bound quadratic in small anisotropy. We also bound the imbalance from random preference patterns and from averaging independent tasks with an isotropic population covariance. With known Λ, changing one term of the spectral contrastive loss selects task-optimal features; with a labelled task bank, we propose a diagnostic and test correction without retraining. Correction reduces synthetic held-out regret from $0.858$ to $0.003$ with $2000$ tasks when less task-relevant directions dominate, and on CIFAR-100 lowers empirical regret by $0.031−−0.053$ on held-out fine-label tasks using the same images. With a small task bank, however, correction can hurt.
ER -