AI Paper Atlas

Research map / Learning theory and algorithms / Matrix and tensor methods

Machine Learning With Tree Tensor Networks, CP Rank Constraints, and Tensor Dropout

Hao Chen, Thomas Barthel

TPAMI 2024 · Learning theory and algorithms › Matrix and tensor methods · approach: Tensor network decompositions · 32 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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 CP rank constraint (low-rank tensors) to the TTN classifier (ablation)TTN classifier, b=4 · MNISTaccuracy (%): 92-93% validation (full TTN, m>4) -> 98.3% testImproved
add CP rank constraint (low-rank tensors) to the TTN classifier (ablation)TTN classifier, b=4 · Fashion-MNISTaccuracy (%): 84% validation (full TTN) -> 88.5% testImproved
add tensor dropout (p=0.2) to low-rank TTN (ablation)low-rank TTN, b=4 · Fashion-MNISTvalidation accuracy: improved by about 1-2%Improved
add tensor dropout to low-rank TTN (best at m=16, r=64, p=0.3) (ablation)low-rank TTN, b=4 · Fashion-MNIST testaccuracy (%): 88.5% -> 90.3%Improved
increase branching ratio from 2 to 4 (ablation)low-rank TTN · Fashion-MNIST testaccuracy (%): b=2 generally lower by around 4%Improved
increase bond dimension m in full TTN (ablation)full TTN, b=4 · MNISTvalidation accuracy: m > 4 all give approximately 92-93%No clear effect
increase CP rank r beyond bond dimension m (ablation)low-rank TTN · MNIST and Fashion-MNISTaccuracyNo clear effect
increase tensor dropout rate beyond critical p_c (ablation)low-rank TTN, b=4, m=16 · Fashion-MNIST testaccuracyMixed
low-rank TTN with dropout versus AlexNetlow-rank TTN · Fashion-MNIST testaccuracy (%): 89.9% -> 90.3%Improved
low-rank TTN with dropout versus GoogLeNetlow-rank TTN · Fashion-MNIST testaccuracy (%): 93.7% vs 90.3%Hurt

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