Towards General Robustness Verification of MaxPool-Based Convolutional Neural Networks via Tightening Linear Approximation | IEEE Conference Publication | IEEE Xplore