DeepPrism: Channel Convolution for Lightweight Generative Models
Open Access
- 24 November 2023
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 178-185
Abstract
In this paper, we introduce DeepPrism, a novel network architecture for generative models, which addresses the redundancy issue of convolutional filters. DeepPrism reaches an unprecedented parameter efficiency improvement of up to 1000 times on generative models. The number of parameters is reduced over conventional CNNs from quadratic to constant dependency with respect to the network's width. The training / inference time and space are also reduced. The main novelty lies in the geometric property namely the translation equivariance on channels, which gives rise to the convolution structure along the channels, and trivially generalizes to the attention mechanism. Compared with Latent Diffusion Model, DeepPrism produces similar qualitative and quantitative results, but the number of parameters can be reduced at a great scale. Other generative models including autoencoders and single-image diffusion models are also experimented to exhibit the generality of this method on generative models.Keywords
Funding Information
- Google TRC Program
- PRAIRIE Institute (ANR-19-P3IA-0001)
- GENCI (2022-AD011013178)
This publication has 20 references indexed in Scilit:
- Structure and Content-Guided Video Synthesis with Diffusion Models2023
- A Comprehensive Survey of Neural Architecture SearchACM Computing Surveys, 2021
- Generative adversarial networksCommunications of the ACM, 2020
- Dense Steerable Filter CNNs for Exploiting Rotational Symmetry in Histology ImagesIEEE Transactions on Medical Imaging, 2020
- Sigmoid-weighted linear units for neural network function approximation in reinforcement learningNeural Networks, 2018
- Group Normalization2018
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric2018
- ImageNet classification with deep convolutional neural networksCommunications of the ACM, 2017
- Deep Residual Learning for Image Recognition2016
- On the Optimality of Shape and Data Representation in the Spectral DomainSIAM Journal on Imaging Sciences, 2015