Stochastic downsampling for cost-adjustable inference and improved regularization in convolutional networks

• Main Authors: Kuen, Jason, Kong, Xiangfei, Lin, Zhe, Wang, Gang, Yin, Jianxiong, See, Simon, Tan, Yap-Peng
• Other Authors: School of Electrical and Electronic Engineering
• Format: Conference or Workshop Item
• Language: English
• Published: 2020
• Subjects: Engineering::Electrical and electronic engineering; Network Parameters; Neural Nets
• Online Access: https://hdl.handle.net/10356/143626
• Institution: Nanyang Technological University

It is desirable to train convolutional networks (CNNs) to run more efficiently during inference. In many cases however, the computational budget that the system has for inference cannot be known beforehand during training, or the inference budget is dependent on the changing real-time resource avail...