State-of-the-art neural networks are getting deeper and wider. While their performance increases with the increasing number of layers and neurons, it is crucial to design an efficient deep architecture in order to reduce computational and memory costs. Designing an efficient neural network, however, is labor intensive requiring many experiments, and fine-tunings. In this paper, we introduce network trimming which iteratively optimizes the network by pruning unimportant neurons based on analysis of their outputs on a large dataset. Our algorithm is inspired by an observation that the outputs of a significant portion of neurons in a large network are mostly zero, regardless of what inputs the network received. These zero activation neurons are redundant, and can be removed without affecting the overall accuracy of the network. After pruning the zero activation neurons, we retrain the network using the weights before pruning as initialization. We alternate the pruning and retraining to further reduce zero activations in a network. Our experiments on the LeNet and VGG-16 show that we can achieve high compression ratio of parameters without losing or even achieving higher accuracy than the original network. * Part of the work was done when Hengyuan Hu and Rui Peng were interns in SenseTime Group Limited
Transient materials capable of disappearing rapidly and completely are critical for transient electronics. End-capped polyoxymethylene (POM) has excellent mechanical properties and thermal stability. However, research concerning the inherent thermal instability of POM without end-capping to obtain transient rather than stable materials, has never been reported. Here, POM without end-capping is proposed as a novel thermally triggered transient solid material that can vanish rapidly by undergoing conversion to a volatile gas, and a chemical vapor deposition method is developed to obtain a smooth POM substrate from the synthesized POM powder. Experimental and theoretical analysis was employed to reveal the mechanism whereby the POM substrate formed and vanished. A Cr/Au/SiO2/Cu memristor device, which was successfully deposited on the POM substrate by physical vapor deposition, exhibits bipolar resistive switching, suggesting that the POM substrate is suitable for use in electrical devices. Thermal triggering causes the POM substrate to vanish as the memristor disintegrates, confirming excellent transient performance. The deposited bulk POM material can completely vanish by thermally triggered depolymerization, and is suitable for physically transient substrates and packaging materials, demonstrating great prospects for application in transient electronics for information security.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.