Presentation | 2019-03-06 A study of inner feature continuity of the VGG model Toya Teramoto, Hyaru Shouno, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Deep Convolutional Neural Network (DCNN) is a successful model in the field of computer vision such like image classification and segmentation and so on. DCNN is a hierarchical neural network model based on Neocognitron, which is inspired from mammalian visual sysytem. There is a VGG model as a popular architecture of DCNN.The VGG block is simple module that performs pooling processing after repeating convolutional layer. Suzuki & Shouno shows an analysis of a DCNN architecture called Network in Network (NIN). They sorted the feature map along the weight strength and found similar feature representation is tend to gather. In this study, we apply their method into the VGG network.We focus the convolution layers in VGG block which can be regarded as a extension of NIN block, so that we analyze the VGG block in the same manner with Suzuki & Shouno. As a result, it was found that multiple convolution processes in the VGG block work to accumulate filters that respond to similar features, and are related to azimuth creativity in the visual cortex. These results suggest that the biological structure is useful for pattern recognition. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | DCNN / Neocognitron / VGG |
Paper # | NC2018-88 |
Date of Issue | 2019-02-25 (NC) |
Conference Information | |
Committee | NC / MBE |
---|---|
Conference Date | 2019/3/4(3days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | University of Electro Communications |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | Yutaka Hirata(Chubu Univ.) / Masaki Kyoso(TCU) |
Vice Chair | Hayaru Shouno(UEC) / Taishin Nomura(Osaka Univ.) |
Secretary | Hayaru Shouno(Nagoya Univ.) / Taishin Nomura(NAIST) |
Assistant | Keiichiro Inagaki(Chubu Univ.) / Takashi Shinozaki(NICT) / Takumi Kobayashi(YNU) / Yasuyuki Suzuki(Osaka Univ.) |
Paper Information | |
Registration To | Technical Committee on Neurocomputing / Technical Committee on ME and Bio Cybernetics |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | A study of inner feature continuity of the VGG model |
Sub Title (in English) | |
Keyword(1) | DCNN |
Keyword(2) | Neocognitron |
Keyword(3) | VGG |
1st Author's Name | Toya Teramoto |
1st Author's Affiliation | University of Electro-Communications(UEC) |
2nd Author's Name | Hyaru Shouno |
2nd Author's Affiliation | University of Electro-Communications(UEC) |
Date | 2019-03-06 |
Paper # | NC2018-88 |
Volume (vol) | vol.118 |
Number (no) | NC-470 |
Page | pp.pp.239-244(NC), |
#Pages | 6 |
Date of Issue | 2019-02-25 (NC) |