no code implementations • 1 Jan 2021 • Alex Renda, Joshua Wolff Fromm
Binarized neural networks, networks with weights and activations constrained to lie in a 2-element set, allow for more time- and resource-efficient inference than standard floating-point networks.
