BEND: Bagging Deep Learning Training Based on Efficient Neural Network Diffusion
Core Concepts
Proposing BEND, a novel approach using diffusion models for efficient deep learning training and inference in Bagging.
Abstract
Bagging integrates multiple base classifiers to reduce model variance.
Traditional deep learning training methods are costly and challenging for diverse models.
Diffusion models efficiently generate neural network parameters with diversity.
BEND utilizes neural network diffusion for building base classifiers in Bagging.
The approach is simple yet effective, generating diverse base classifiers for various tasks.
Experimental results show BEND outperforms original and diffused models consistently.
"Diffusion models efficiently generate neural network parameters with diversity."
"Our proposed BEND algorithm consistently outperforms the mean and median accuracies of both the original trained model and the diffused model."