Margin-Maximization in Binarized Neural Networks for Optimizing Bit Error Tolerance | IEEE Conference Publication | IEEE Xplore