[CapsNet]基于pytorch的胶囊网络工具包

最新推荐文章于 2024-06-05 09:56:23 发布

身披白袍

最新推荐文章于 2024-06-05 09:56:23 发布

阅读量1.2k

点赞数 1

分类专栏：机器学习|数据挖掘|数学模型文章标签：胶囊网络 CapsNet Pytorch 工具包

本文链接：https://blog.csdn.net/Shenpibaipao/article/details/103476142

版权

API

论文地址：Dynamic Routing Between Capsules

开发了一个基于pytorch的胶囊网络工具包，以下是它的API：

### 胶囊化层
class Capsulation2D(nn.Module)
	# input.shape = (batch_size, channels, height, weight)
    # output.shape = (batch_size, out_channels, out_dim_capsule, height, weight)
    
### 反胶囊化层
class DeCapsulation2D(nn.Module)
	# input.shape = (batch_size, channels, dim_capsule, height, weight)
    # output.shape = (batch_size, out_channels, height, weight)
    
### 平化层
class CapFlatten(nn.Module)
    # input.shape = (batch_size, channels, dim_capsule, height, weight)
    # output.shape = (batch_size, channels * height * weight, dim_capsule) which is (batch_size, num_capsules, dim_capsule)


### 反平化层