"Learning-Compression" Algorithms for Neural Net Pruning

• Published in: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition pp. 8532 - 8541
• Main Authors: Carreira-Perpinan, Miguel A., Idelbayev, Yerlan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2018
• Subjects: Mobile handsets; Neural networks; Neurons; Optimization; Performance evaluation; Quantization (signal); Training

Pruning a neural net consists of removing weights without degrading its performance. This is an old problem of renewed interest because of the need to compress ever larger nets so they can run in mobile devices. Pruning has been traditionally done by ranking or penalizing weights according to some c...