Boosting Certified ℓ_∞ Robustness with EMA Method and Ensemble Model
share
The neural network with 1-Lipschitz property based on ℓ_∞-dist neuron has a theoretical guarantee in certified ℓ_∞ robustness. However, due to the inherent difficulties in the training of the network, the certified accuracy of previous work is limited. In this paper, we propose two approaches to deal with these difficuties. Aiming at the characteristics of the training process based on ℓ_∞-norm neural network, we introduce the EMA method to improve the training process. Considering the randomness of the training algorithm, we propose an ensemble method based on trained base models that have the 1-Lipschitz property and gain significant improvement in the small parameter network. Moreover, we give the theoretical analysis of the ensemble method based on the 1-Lipschitz property on the certified robustness, which ensures the effectiveness and stability of the algorithm. Our code is available at https://github.com/Theia-4869/EMA-and-Ensemble-Lip-Networks.
READ FULL TEXT