Fast Ensembling with Diffusion Schr\"odinger Bridge

arXiv:2404.15814v1 Announce Type: new
Abstract: Deep Ensemble (DE) approach is a straightforward technique used to enhance the performance of deep neural networks by training them from different initial points, converging towards various local optima. However, a limitation of this methodology lies in its high computational overhead for inference, arising from the necessity to store numerous learned parameters and execute individual forward passes for each parameter during the inference stage. We propose a novel approach called Diffusion Bridge Network (DBN) to …

arxiv bridge cs.ai cs.lg diffusion type