Alternating direction method of multipliers for nonlinear image restoration problems.
PulseAugur coverage of Alternating direction method of multipliers for nonlinear image restoration problems. — every cluster mentioning Alternating direction method of multipliers for nonlinear image restoration problems. across labs, papers, and developer communities, ranked by signal.
8 day(s) with sentiment data
-
New NRSfM solver drastically cuts computation time while maintaining accuracy
Researchers have developed a novel solver for Non-rigid Structure from Motion (NRSfM) that significantly improves speed and scalability. This new approach bypasses computationally intensive steps like semidefinite progr…
-
New Convex Estimator for Robust Gaussian Graphical Models
Researchers have developed a novel convex estimator for robustly identifying Gaussian graphical models, particularly those with a sparse component combined with a low-rank matrix. This method utilizes a Gaussian measure…
-
OPTIMA method offers efficient LLM pruning via quadratic programming
Researchers have developed OPTIMA, a novel one-shot pruning method for large language models that balances accuracy and computational feasibility. This method frames layer-wise weight reconstruction as independent, colu…
-
New SC-RCCTV Model Enhances Hyperspectral Image Restoration
Researchers have developed a novel Self-Calibrated Subspace Robust PCA (SC-RCCTV) model designed to remove mixed noise and line stripes from hyperspectral images without requiring manual parameter tuning. This model ada…
-
New Manifold Deep Unfolding PCA method enhances unsupervised feature selection
Researchers have developed a new unsupervised feature selection method called Manifold Deep Unfolding PCA (MDUFS-Net). This approach integrates the $\ell_{2,0}$-norm into principal component analysis and unfolds an ADMM…
-
New CACTUS optimizer enhances LLM pruning by preserving influence functions
Researchers have developed CACTUS, a novel optimizer that uses a curvature-aware objective to improve post-training pruning of large language models. This method aims to preserve the internal gradient alignment between …
-
New framework offers private and robust federated multiclass classification
Researchers have developed a new framework for federated multiclass classification that offers both differential privacy and distributional robustness. This approach enables collaborative training of models across multi…
-
HUANet: New deep learning architecture accelerates constrained convex optimization
Researchers have introduced HUANet, a novel deep neural network architecture designed to accelerate constrained convex optimization. Unlike previous black-box methods, HUANet explicitly incorporates optimality principle…
-
New ShamAN-Q method slashes LLM weights to sub-1-bit
Researchers have developed ShamAN-Q, a novel sub-1-bit post-training quantization method for large language models. This technique enhances NanoQuant by incorporating a dense curvature metric derived from the Shampoo op…
-
New tensor decomposition methods improve image recovery and data completion · 2 sources tracked
Two new research papers propose novel methods for low-rank tensor completion and multi-dimensional image recovery. The first paper introduces two weighted Schatten-p tensor factorization models, WSpTFI and WSpTFII, desi…
-
New research unifies linear recommendation models under norm-based regularization
This paper investigates the regularization landscape of linear recommendation models, finding that top-performing models primarily use either nuclear-norm or Frobenius-norm based regularizers. While nuclear-norm solutio…
-
New method efficiently learns balanced signed graphs using sparse linear programming
Researchers have developed a novel method for efficiently learning balanced signed graphs, which incorporate both positive and negative correlations in data. This new approach extends a linear programming-based techniqu…
-
New taxonomy classifies non-convex optimization regimes using Lagrange multipliers
A new research paper introduces a taxonomy for non-convex optimization problems by analyzing the signature of Lagrange multipliers at KKT stationary points. The taxonomy categorizes problems into five operational regime…
-
New ADMM-Q algorithm enhances LLM quantization, reducing perplexity
Researchers have developed ADMM-Q, a new algorithm designed to improve post-training quantization for large language models. This method utilizes a combinatorial variant of the Alternating Direction Method of Multiplier…
-
New SNMF Method Enhances Sparsity and Identifiability with Novel Regularization
Researchers have developed a new method for Separable Nonnegative Matrix Factorization (SNMF) that enhances sparsity and identifiability of learned factors. This approach utilizes a powered ratio-of-norms regularizer, l…
-
New CM-GLasso framework learns interpretable visual-linguistic dependency graphs
Researchers have developed CM-GLasso, a novel framework for learning interpretable conditional-dependence structures from multimodal visual-linguistic data. This approach integrates vision-language representation learni…
-
Frozen CLIP Priors Enhance Self-Supervised Imaging for Poisson Noise
Researchers have developed a novel self-supervised learning method for imaging inverse problems, particularly effective under Poisson noise. This approach utilizes frozen CLIP RN50 features as a parameter-efficient prio…
-
New MS-WDRO framework fuses heterogeneous graph data using Wasserstein metric
Researchers have developed a novel framework called MS-WDRO for learning graph structures from multiple, heterogeneous data sources. This method leverages the Wasserstein metric to fuse diverse datasets by calculating a…
-
AI helps discover counterexample to ADMM convergence
Researchers have discovered a counterexample to the convergence of a specific type of three-block Alternating Direction Method of Multipliers (ADMM) when the third constraint block is an identity matrix. Utilizing AI to…
-
New paper explores information bottleneck under perfect privacy
This paper explores the information bottleneck principle under the condition of perfect privacy, focusing on scenarios where the representation-rate constraint is active. The objective is to create a representation that…