Efficient Neural Network Robustness Certification with General Activation Functions
Authors: Huan Zhang, Tsui-Wei Weng, Pin-Yu Chen, Cho-Jui Hsieh, Luca Daniel
NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results show that CROWN on Re LU networks can notably improve the certified lower bounds compared to the current state-of-the-art algorithm Fast-Lin, while having comparable computational efficiency. Furthermore, CROWN also demonstrates its effectiveness and flexibility on networks with general activation functions, including tanh, sigmoid and arctan. |
| Researcher Affiliation | Collaboration | 1University of California, Los Angeles, Los Angeles CA 90095 2Massachusetts Institute of Technology, Cambridge, MA 02139 3MIT-IBM Watson AI Lab, IBM Research, Yorktown Heights, NY 10598 |
| Pseudocode | No | The paper describes the algorithm steps in text and mathematical formulations but does not present pseudocode or a clearly labeled algorithm block. |
| Open Source Code | Yes | Our code is available at https://github.com/CROWN-Robustness/Crown |
| Open Datasets | Yes | We evaluate CROWN and other baselines on multi-layer perceptron (MLP) models trained on MNIST and CIFAR-10 datasets. |
| Dataset Splits | No | The paper mentions using 100 random test images and random attack targets and implies standard use of MNIST and CIFAR-10, but it does not specify explicit train/validation/test splits (e.g., percentages or counts) or reference predefined splits with citations for reproducibility. |
| Hardware Specification | Yes | Experiments were conducted on an Intel Skylake server CPU running at 2.0 GHz on Google Cloud. |
| Software Dependencies | No | We implement our algorithm using Python (numpy with numba). Most computations in our method are matrix operations that can be automatically parallelized by the BLAS library; however, we set the number of BLAS threads to 1 for a fair comparison to other methods. The paper mentions Python, numpy, and numba, but does not provide specific version numbers for these software components, nor for the BLAS library. |
| Experiment Setup | No | The paper mentions searching for 'best learning rate and weight decay parameters' when training NN models with non-ReLU activation functions, but it does not explicitly provide the specific values for these or other hyperparameters (e.g., batch size, number of epochs, optimizer settings) that were used in the reported experiments. |