Timezone: »
Filters in a Convolutional Neural Network (CNN) contain model parameters learned from enormous amounts of data. In this paper, we suggest to decompose convolutional filters in CNN as a truncated expansion with pre-fixed bases, namely the Decomposed Convolutional Filters network (DCFNet), where the expansion coefficients remain learned from data. Such a structure not only reduces the number of trainable parameters and computation, but also imposes filter regularity by bases truncation. Through extensive experiments, we consistently observe that DCFNet maintains accuracy for image classification tasks with a significant reduction of model parameters, particularly with Fourier-Bessel (FB) bases, and even with random bases. Theoretically, we analyze the representation stability of DCFNet with respect to input variations, and prove representation stability under generic assumptions on the expansion coefficients. The analysis is consistent with the empirical observations.
Author Information
Qiang Qiu (Duke University)
Xiuyuan Cheng (Duke University)
robert Calderbank (Duke University)
Guillermo Sapiro (Duke University)
Related Events (a corresponding poster, oral, or spotlight)
-
2018 Oral: DCFNet: Deep Neural Network with Decomposed Convolutional Filters »
Fri Jul 13th 03:30 -- 03:40 PM Room K1
More from the Same Authors
-
2019 Poster: Adversarially Learned Representations for Information Obfuscation and Inference »
Martin A Bertran · Natalia Martinez · Afroditi Papadaki · Qiang Qiu · Miguel Rodrigues · Galen Reeves · Guillermo Sapiro -
2019 Oral: Adversarially Learned Representations for Information Obfuscation and Inference »
Martin A Bertran · Natalia Martinez · Afroditi Papadaki · Qiang Qiu · Miguel Rodrigues · Galen Reeves · Guillermo Sapiro