We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
No description provided.
The text was updated successfully, but these errors were encountered:
Paper 第四页有详细说明,这篇博客中也有一些介绍。我对这个领域不熟悉,我的理解是这样的:作者想要让相同类型的capsule处理相同类型的entity,但同时又想要保留位置的信息,使得不同位置output的值不同,所以把坐标加进vote matrix。
我觉得也可以类比于conv来理解,比如一张image上有4个车轮,那这4个车轮可以用同样的conv来提取特征,这是卷积核的局部共享。capsule在最后一层的时候,既想要有这种共享的transformation,同时也想encode位置信息,所以采用了Coordinate Addition的技术。作者在Table 1的实验表明(对比Row.3和Row.6),使用这个技术的效果是有提升的。
这是我的理解,欢迎讨论。
Sorry, something went wrong.
No branches or pull requests
No description provided.
The text was updated successfully, but these errors were encountered: