# Orthogonal Convolutional Neural Networks

143 篇文章 6 订阅

#### 文章目录

Wang J, Chen Y, Chakraborty R, et al. Orthogonal Convolutional Neural Networks.[J]. arXiv: Computer Vision and Pattern Recognition, 2019.

@article{wang2019orthogonal,
title={Orthogonal Convolutional Neural Networks.},
author={Wang, Jiayun and Chen, Yubei and Chakraborty, Rudrasis and Yu, Stella X},
journal={arXiv: Computer Vision and Pattern Recognition},
year={2019}}

### 主要内容

#### 符号说明

X ∈ R N × C × H × W X \in \mathbb{R}^{N \times C \times H \times W} : 输入
K ∈ R M × C × k × k K \in \mathbb{R}^{M \times C \times k \times k} : 卷积核
Y ∈ R N × M × H ′ × W ′ Y \in \mathbb{R}^{N \times M \times H' \times W'} : 输出
Y = C o n v ( K , X ) Y= Conv(K,X)

#### kernel orthogonal regularization

L k o r t h − r o w = ∥ K K T − I ∥ F , L k o r t h − c o l = ∥ K T K − I ∥ F . L_{korth-row}= \|KK^T-I\|_F,\\ L_{korth-col}= \|K^TK-I\|_F.

#### orthogonal convolution

K ( i h w , ⋅ ) \mathcal{K}(ihw,\cdot) 表示第 ( i − 1 ) H ′ W ′ + ( h − 1 ) W ′ + w (i-1) H'W'+(h-1)W'+w 行, 对应的 K ( ⋅ , i h w ) \mathcal{K}(\cdot, ihw) 表示 ( i − 1 ) H W + ( h − 1 ) W + w (i-1) HW+(h-1)W+w 列.

K K T = I \mathcal{K}\mathcal{K}^T=I 等价于
⟨ K ( i h 1 w 1 , ⋅ ) , K ( j h 2 w 2 , ⋅ ) ⟩ = { 1 , ( i , h 1 , w 1 ) = ( j , h 2 , w 2 ) 0 , e l s e . (5) \tag{5} \langle \mathcal{K}(ih_1w_1, \cdot), \mathcal{K}(jh_2w_2,\cdot)\rangle = \left \{ \begin{array}{ll} 1, & (i,h_1,w_1)=(j,h_2,w_2) \\ 0, & else. \end{array} \right.
K T K = I \mathcal{K}^T\mathcal{K}=I 等价于
⟨ K ( ⋅ , i h 1 w 1 ) , K ( ⋅ , j h 2 w 2 ) ⟩ = { 1 , ( i , h 1 , w 1 ) = ( j , h 2 , w 2 ) 0 , e l s e . (10) \tag{10} \langle \mathcal{K}(\cdot, ih_1w_1), \mathcal{K}(\cdot, jh_2w_2)\rangle = \left \{ \begin{array}{ll} 1, & (i,h_1,w_1)=(j,h_2,w_2) \\ 0, & else. \end{array} \right.

(5)等价于
C o n v ( K , K , p a d d i n g = P , s t r i d e = S ) = I r 0 , (7) \tag{7} Conv(K, K,padding=P, stride=S)=I_{r0},

P = ⌊ k − 1 S ⌋ ⋅ S . P= \lfloor \frac{k-1}{S} \rfloor \cdot S.

K T K \mathcal{K}^T\mathcal{K} S = 1 S=1 特殊情况下的特殊情况下, (10)等价于
C o n v ( K T , K T , p a d d i n g = k − 1 , s t r i d e = 1 ) = I c 0 , (11) \tag{11} Conv (K^T,K^T, padding=k-1, stride=1)=I_{c0},

min ⁡ K ∥ K K T − I ∥ F \min_K \|\mathcal{K}\mathcal{K}^T-I\|_F

min ⁡ K ∥ K T K − I ∥ F \min_K \|\mathcal{K}^T\mathcal{K}-I\|_F

C o n v ( K , K , p a d d i n g = 0 ) = I r 0 C o n v ( K T , K T , p a d d i n g = 0 ) = I c 0 , Conv(K,K,padding=0)=I_{r0} \\ Conv(K^T, K^T, padding=0)=I_{c_0},

• 2
点赞
• 2
收藏
觉得还不错? 一键收藏
• 打赏
• 2
评论
09-18 725
03-04 1065
08-02 289
04-10 66万+
01-23 5795
07-26 2万+

### “相关推荐”对你有帮助么？

• 非常没帮助
• 没帮助
• 一般
• 有帮助
• 非常有帮助

MTandHJ

¥1 ¥2 ¥4 ¥6 ¥10 ¥20

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。