Year: 2019
Type: preprint
Abstract: Due to their high computational complexity, deep neural networks are still limited to powerful processing units. To promote a reduced model complexity by dint of low-bit fixed-point quantization, we propose a gradient-based optimization strategy to generate a symmetric mixture of Gaussian modes (SGM) where each mode belongs to a particular quantization stage. We achieve 2-bit state-of-the-art performance and illustrate the model's ability for self-dependent weight adaptation during training.
Source: arXiv (Cornell University)
Authors: Enderich, Lukas, Timm, Fabian, Rosenbaum, Lars, Burgard, Wolfram
Institutions: Robert Bosch (Germany), University of Freiburg
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2961372437