Low-Precision Floating-Point for Efficient On-Board Deep Neural Network Processing | IEEE Conference Publication | IEEE Xplore