Binghui Li

collaborators

2 papers

cs.CV2022

Reconstruction Task Finds Universal Winning Tickets

Ruichen Li, Binghui Li, Qi Qian +1

Pruning well-trained neural networks is effective to achieve a promising accuracy-efficiency trade-off in computer vision regimes. However, most of existing pruning algorithms only…

cs.LG2021

Boosting Certified $ℓ_{\infty}$ Robustness with EMA Method and Ensemble Model

Binghui Li, Shiji Xin, Qizhe Zhang

The neural network with $1$ -Lipschitz property based on $ℓ_{\infty}$ -dist neuron has a theoretical guarantee in certified $ℓ_{\infty}$ robustness. However, due to the inherent di…

Reconstruction Task Finds Universal Winning Tickets

Boosting Certified ℓ∞​ Robustness with EMA Method and Ensemble Model

Boosting Certified $ℓ_{\infty}$ Robustness with EMA Method and Ensemble Model