AI Paper Atlas

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

DeepTensor: Low-Rank Tensor Decomposition With Deep Network Priors

Vishwanath Saragadam, Randall Balestriero, Ashok Veeraraghavan, Richard G. Baraniuk

TPAMI 2024 · Learning theory and algorithms › Matrix and tensor methods · approach: Geometry of low-rank and SPD structure · 56 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
use fixed schedule instead of cosine annealing at max LR 1e-2 (ablation)DeepTensor · toy matrix low-rank approximationPSNR (dB): 16.5 -> 17.6Improved
use max learning rate 1e-4 or higher instead of 1e-5 (fixed schedule) (ablation)DeepTensor · toy matrix low-rank approximationPSNR (dB): 16.5 -> 17.6Improved
replace SVD with DeepTensor (ablation)SVD · toy matrix low-rank approximationPSNR (dB): 14.4 -> 17.6Improved
use ReLU as final activation instead of softplus (ablation)DeepTensor NMF · MNIST NMFPSNR (dB): 7.4 -> 7.6Improved
use ReLU as final activation instead of softplus (ablation)DeepTensor NMF · CIFAR10 NMFPSNR (dB): 8.4 -> 8.8Improved
replace classical NMF with DeepTensor (ReLU) (ablation)NMF · MNIST NMFPSNR (dB): 7.5 -> 7.6Improved
replace classical NMF with DeepTensor (ReLU)NMF · CIFAR10 NMFPSNR (dB): 8.2 -> 8.8Improved
replace baseline NMF with DeepTensorNMF · GoogleCommands spectrograms NMFPSNR (dB): 4.6 -> 9.0Improved
replace baseline semi-NMF with DeepTensorNMF · GoogleCommands spectrograms semi-NMFPSNR (dB): 4.7 -> 8.8Improved
replace SVD with DeepTensor under Poisson noiseSVD · hyperspectral denoising (Poisson, rank 20)PSNR (dB): approximately 6dB increaseImproved
replace BM3D with DeepTensor under Poisson noiseBM3D · hyperspectral denoising (Poisson, rank 20)PSNR (dB): 3dB over BM3DImproved
use DeepTensor instead of single 3D-conv DN for tensor decompositionsingle DN with 3D convolutions · 3D tensor decompositiontraining time: 60 times fasterImproved

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