Pith. sign in

REVIEW 6 cited by

Learning from Similar Linear Representations: Adaptivity, Minimaxity, and Robustness

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2303.17765 v5 pith:KPCFQET4 submitted 2023-03-31 stat.ML cs.LG

classification stat.MLcs.LG
keywords taskslearningoutlierrepresentationrepresentationstextitmethodsame
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Representation multi-task learning (MTL) has achieved tremendous success in practice. However, the theoretical understanding of these methods is still lacking. Most existing theoretical works focus on cases where all tasks share the same representation, and claim that MTL almost always improves performance. Nevertheless, as the number of tasks grows, assuming all tasks share the same representation is unrealistic. Furthermore, empirical findings often indicate that a shared representation does not necessarily improve single-task learning performance. In this paper, we aim to understand how to learn from tasks with \textit{similar but not exactly the same} linear representations, while dealing with outlier tasks. Assuming a known intrinsic dimension, we propose a penalized empirical risk minimization method and a spectral method that are \textit{adaptive} to the similarity structure and \textit{robust} to outlier tasks. Both algorithms outperform single-task learning when representations across tasks are sufficiently similar and the proportion of outlier tasks is small. Moreover, they always perform at least as well as single-task learning, even when the representations are dissimilar. We provide information-theoretic lower bounds to demonstrate that both methods are nearly \textit{minimax} optimal in a large regime, with the spectral method being optimal in the absence of outlier tasks. Additionally, we introduce a thresholding algorithm to adapt to an unknown intrinsic dimension. We conduct extensive numerical experiments to validate our theoretical findings.

Discussion (0). Sign in to comment.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Contaminated Multi-task Learning with Heterogeneity: Fundamental Limits and Optimal Algorithms

    stat.ML 2026-07 accept novelty 7.5 of 10

    Filtering-based robust multi-task gradient descent matches minimax rates under task contamination and heterogeneity, removing the √d contamination barrier of regularization and score-based methods.

  2. On Non-Stationary Dynamic Pricing: Adaptivity and Optimality

    stat.ML 2026-07 conditional novelty 7.0 of 10

    An adaptive dynamic-pricing algorithm achieves, up to logarithmic factors, the minimax optimal regret for both abrupt and smooth non-stationarity in contextual GLM demand, and comes with a matching lower bound.

  3. Representation-Aware Distributionally Robust Optimization: A Knowledge Transfer Framework

    stat.ME 2025-09 conditional novelty 6.0 of 10

    A representation-aware Wasserstein DRO framework that shrinks estimators toward an external representation subspace, with asymptotic inference and adaptive robustness tuning.

  4. Late Fusion Multi-task Learning for Semiparametric Inference with Nuisance Parameters

    stat.ME 2025-07 conditional novelty 6.0 of 10

    A late-fusion multi-task learning framework for double machine learning, with theories showing faster rates when tasks share similar parameters, plus a fused kernel method for nuisance parameters.

  5. Joint estimation of smooth graph signals from partial linear measurements

    math.ST 2025-05 conditional novelty 6.0 of 10

    A smoothness-penalized least squares estimator jointly recovers graph-indexed signals from partial noisy measurements, and is weakly consistent as the graph grows even when per-vertex measurements are rank-one and mos...

  6. Multi-task Learning for Heterogeneous Multi-source Block-Wise Missing Data

    cs.LG 2025-05 conditional novelty 5.0 of 10

    A two-step framework combining block-wise imputation with shared and task-specific multi-task learning improves prediction under distribution and posterior shift.

Pith tools