Conference paper
Training deep neural networks with 8-bit floating point numbers
Naigang Wang, Jungwook Choi, et al.
NeurIPS 2018
Paper
On typing in prolog
Abstract
Frequently a domain of discourse can be described as a hierarchical structure. Then instead of using one place predicates to describe the type of a term one can associate types directly, and let the unification procedure take advantage of it. This prunes the search space without the use of the Cut operator. © 1986, ACM. All rights reserved.
Related
Ewout van den Berg, Daniel Brand, et al.
INTERSPEECH 2015
Conference paper
MEC: Memory-efficient convolution for deep neural network
M. Cho, Daniel Brand
ICML 2017
Daniel Brand, Mareio Buss, et al.
ICSM 2007