未验证 提交 2c1270e4 编写于 作者: K kexinzhao 提交者: GitHub

fix maxout op latex equation (#6303)

上级 458ffbf4
...@@ -40,23 +40,28 @@ class MaxOutOpMaker : public framework::OpProtoAndCheckerMaker { ...@@ -40,23 +40,28 @@ class MaxOutOpMaker : public framework::OpProtoAndCheckerMaker {
"the number of channels divided by groups.." "the number of channels divided by groups.."
)DOC"); )DOC");
AddComment(R"DOC( AddComment(R"DOC(
Assumed the input shape is (N, Ci, H, W). MaxOut Operator.
The output shape is (N, Co, H, W). Then `Co = Ci / groups`.
math: Assumed the input shape is (N, Ci, H, W).
y_{si+j} = \max_k x_{gsi + sk + j} The output shape is (N, Co, H, W).
g = groups Then $Co = Ci / groups$ and the operator formula is as follows:
s = input.size / num_channels
0 \le i < num_channels / groups
0 \le j < s
0 \le k < groups
Please refer to Paper: $$
y_{si+j} = \max_k x_{gsi + sk + j} \\
g = groups \\
s = \frac{input.size}{num\_channels} \\
0 \le i < \frac{num\_channels}{groups} \\
0 \le j < s \\
0 \le k < groups
$$
Please refer to Paper:
- Maxout Networks: http://www.jmlr.org/proceedings/papers/v28/goodfellow13.pdf - Maxout Networks: http://www.jmlr.org/proceedings/papers/v28/goodfellow13.pdf
- Multi-digit Number Recognition from Street View \ - Multi-digit Number Recognition from Street View \
Imagery using Deep Convolutional Neural Networks: \ Imagery using Deep Convolutional Neural Networks: \
https://arxiv.org/pdf/1312.6082v4.pdf https://arxiv.org/pdf/1312.6082v4.pdf
)DOC");
)DOC");
} }
}; };
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册