Algorithm-Hardware Co-Design of Adaptive Floating-Point Encodings for Resilient Deep Learning Inference | IEEE Conference Publication | IEEE Xplore