Latest Articles for Neural Networks

Introducing a novel optimization framework for training neural networks using manifold structures.

This article highlights certified training methods for robust neural networks.

New methods for boosting SNN resistance against attacks.

A fresh approach to improving Graph Neural Networks using model soups.

A look at neural operators, focusing on injectivity and bijectivity.

SMNNs combine topology and neural networks for improved data processing and explainability.

SeqBoat utilizes Sparse Modular Activation for efficient sequence processing.

A look at prompt-tuning and its impact on attention in language models.

New model mimics biological patterns using neural networks without external input.

A new model enhances neural network representation learning using GNN and Transformer features.

This article discusses partial hypernetworks as a solution for catastrophic forgetting.

New adaptive methods improve bi-level optimization in deep learning.

Progressive Fourier Neural Representation enhances video processing while retaining quality.

A framework to merge traditional and modern dimensionality reduction methods for model interpretability.

A new approach enhances learning in partially observable environments.

New methods improve efficiency and accuracy in training transformers with low-precision arithmetic.

A new approach enhances pooling configurations in convolutional neural networks.

A new approach to train models effectively despite noisy labels.

New methods tackle challenges in neural networks using memristor technology.

A new method enhances Class Incremental Learning performance and reduces forgetting.

Research highlights the benefits of the reducing-width Ansatz design in QNNs.

A new approach to verify neural networks' global behaviors for reliable performance.

New methods improve translation accuracy while using specific vocabulary.

This article explores grokking and its effects on neural network training.

SingNet improves beat tracking in singing voices using past data.

New methods aim to reduce outliers in transformer models, enhancing quantization efficiency.

Examining the rise of spiking neural networks in artificial intelligence.

Research highlights challenges and solutions for deep learning systems to adapt over time.

VCReg enhances transfer learning by promoting diverse feature representations in models.

New neural network model tackles challenges in analyzing dependent data.

Methods to ensure Bayesian Neural Networks are reliable against adversarial attacks.

NetBooster improves performance of tiny neural networks for smart devices.

A new G-CNN improves 3D medical image analysis and accuracy.

FLOWER addresses few-shot learning and catastrophic forgetting in machine learning models.

Self-encoders enhance data organization and classification in machine learning.

A new method restores damaged frescoes using innovative technology.

A new optimizer promises faster training for deep neural networks.

A fresh approach helps AI learn without forgetting past knowledge.

Innovative approaches tackle inverse problems using learning-based techniques and invertible networks.

A new model improves melody harmonization by considering emotional factors.