Digital Computation-in-Memory Design with Adaptive Floating Point for Deep Neural Networks | IEEE Conference Publication | IEEE Xplore