Proposes a covariance-aware tuning-free shrinkage framework and sequential algorithm for multi-source estimation that attains oracle risk asymptotically and improves on single-step methods.
arXiv preprint arXiv:2005.00944 , year=
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
verdicts
UNVERDICTED 5representative citing papers
Matrix-weighted regularization for robust multi-task regression achieves optimal MSE under weaker spectral assumptions and performs no worse than independent learning when balancedness is poor.
A novel adaptive multi-task learning framework with projection-penalized PCA learns cross-sector factor subspace relatedness to improve multi-sector factor model estimation and portfolio optimization.
iGSP uses implicit gradient subspace projection in two phases to enable efficient continual adaptation of vision-language models, claiming SOTA accuracy with 42.7% fewer trainable parameters and 86.9% less total parameter growth.
SURGELLM introduces a surgical feature gate, task-conditioned prefix tokens, and instance-weighted normalization to transformers, reporting 0.940 macro-F1 across four NLP tasks with code released.
citing papers explorer
-
Tuning-Free Efficient Estimation for Multi-Source Data via Covariance-Aware Shrinkage
Proposes a covariance-aware tuning-free shrinkage framework and sequential algorithm for multi-source estimation that attains oracle risk asymptotically and improves on single-step methods.
-
Multi-task Linear Regression without Eigenvalue Lower Bounds: Adaptivity, Robustness, and Safety
Matrix-weighted regularization for robust multi-task regression achieves optimal MSE under weaker spectral assumptions and performs no worse than independent learning when balancedness is poor.
-
Adaptive Multi-task Learning for Multi-sector Portfolio Optimization
A novel adaptive multi-task learning framework with projection-penalized PCA learns cross-sector factor subspace relatedness to improve multi-sector factor model estimation and portfolio optimization.
-
iGSP:Implicit Gradient Subspace Projection for Efficient Continual Learning of Vision-Language Models
iGSP uses implicit gradient subspace projection in two phases to enable efficient continual adaptation of vision-language models, claiming SOTA accuracy with 42.7% fewer trainable parameters and 86.9% less total parameter growth.
-
SURGELLM: Rethinking Multi-Task Evaluation through Task-Aware Feature Gating with Class-Balanced Normalization
SURGELLM introduces a surgical feature gate, task-conditioned prefix tokens, and instance-weighted normalization to transformers, reporting 0.940 macro-F1 across four NLP tasks with code released.