AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

Sparse GCA and Thresholded Gradient Descent

Sheng Gao, Zongming Ma

JMLR 2023 · Optimization › Implicit bias and generalization · approach: Implicit regularization of SGD · 7 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add thresholded gradient descent refinement after Fantope initializationGeneralized Fantope initial estimator · Sparse GCA 3 datasets, r=1squared matrix distance: 0.1319 -> 0.0015Improved
add thresholded gradient descent refinement after Fantope initializationGeneralized Fantope initial estimator · Sparse GCA 3 datasets, r=5squared matrix distance: 0.1969 -> 0.0171Improved
use TGD instead of Rifle for refinementRifle · Sparse GCA r=1, n=500squared matrix distance: 0.0006 -> 0.0003Improved
use TGD instead of Rifle for refinementRifle · Sparse GCA r=1, n=1000squared matrix distance: 0.0003 -> 0.0002Improved
use TGD instead of CoLaR for refinementCoLaR · Sparse CCA Toeplitz (300,300,200)prediction loss: 0.0511 -> 0.0118Improved
use TGD instead of CoLaR for refinementCoLaR · Sparse CCA Toeplitz (600,600,500)prediction loss: 0.0176 -> 0.0213Hurt
use TGD instead of CoLaR for refinementCoLaR · Sparse CCA SparseInv (300,300,500)prediction loss: 0.0242 -> 0.0042Improved
use TGD instead of CoLaR for refinementCoLaR · Sparse CCA Identity (600,600,500)prediction loss: 0.0166 -> 0.0082Improved
vary penalty lambda from 0.0001 to 1 (ablation)TGD · Sparse GCA r=3squared matrix distance: 0.0121 -> 0.0099No clear effect
vary penalty lambda from 0.0001 to 1 (ablation)TGD · Sparse GCA r=4squared matrix distance: 0.0145 -> 0.0119No clear effect
choose sparsity level s' by 5-fold cross-validation (ablation)TGD · Sparse GCA r=3 (s=15)selected s': 25No clear effect
input latent dimension r=2 when true dimension is 3 (misspecification) (ablation)TGD · Sparse CCA misspecified Identity (300,300,500)prediction loss: 0.0190 -> 0.0113No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy