On the Convergence of Certified Robust Training with Interval Bound Propagation

Authors: Yihan Wang, Zhouxing Shi, Quanquan Gu, Cho-Jui Hsieh

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this paper, we present a theoretical analysis on the convergence of IBP training... We further conduct experiments to compare the convergence of networks with different widths m for natural training and IBP training respectively.
Researcher Affiliation Academia Yihan Wang*, Zhouxing Shi*, Quanquan Gu, Cho-Jui Hsieh University of California, Los Angeles {yihanwang,zshi,qgu,chohsieh}@cs.ucla.edu
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any statement or link regarding the availability of its source code.
Open Datasets Yes We use the MNIST (Le Cun et al., 2010) dataset and take digit images with label 2 and 5 for binary classification.
Dataset Splits No The paper mentions training the model but does not specify training, validation, or test dataset splits.
Hardware Specification Yes even if we enlarge m up to 80,000 limited by the memory of a single Ge Force RTX 2080 GPU
Software Dependencies No The paper does not specify any software dependencies with version numbers.
Experiment Setup Yes We train the model for 70 epochs with SGD, and we keep ϵ fixed throughout the whole training process.