Beyond Flops In Low-Rank Compression Of Neural Networks: Optimizing Device-Specific Inference Runtime | IEEE Conference Publication | IEEE Xplore