Ohyeon5 / pytorch-graphNet

Pytorch implementation of Deepmind's Graph network. Original repo: https://github.com/deepmind/graph_nets.git. Original paper: https://arxiv.org/abs/1806.01261

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

pytorch-graphNet

Current github page is to review Deepmind's Graph network and re-implement it with pytorch. Tutorial will be largely changed and would be explained in Korean.

본 튜토리얼은 Graph network를 pytorch로 구현하고, 네트워크를 예시와 함께 이해하기 쉽도록 한국어로 제작될 예정입니다.

주요 콘텐츠는 다음과 같음

  1. Graph network 의 전반적인 이해와 응용 및 확장 가능성 시사
  2. 리뷰할 논문
  1. Graphnet Shortest-path demo pytorch 구현 & 설명/ relational reasoning 핵심 network dynamics 구현 & 설명

References

Pytorch implementation of Deepmind's Graph network. Original repo: https://github.com/deepmind/graph_nets.git.
Original paper: https://arxiv.org/abs/1806.01261
Battaglia, Peter W., et al. "Relational inductive biases, deep learning, and graph networks." arXiv preprint arXiv:1806.01261 (2018).
Santoro, Adam, et al. "A simple neural network module for relational reasoning." Advances in neural information processing systems. 2017.

What are Graph Networks?

그래프 네트워크는 그래프를 input으로 받고, 그래프를 out으로 받는다. 각 input 그래프는 엣지(E)와 노드(V), 그리고 해당 그래프의 전체적인 속성(global-level attribute)을 변수로 둔다. out 그래프는 동일한 구조를 가지고 있으나 업데이트 된 속성(attribute) 값을 가진다.

Why Graph Networks?

Possible Applications

Demos

  1. Shortest path
  2. Relational reasoning with CNN and NLP

About

Pytorch implementation of Deepmind's Graph network. Original repo: https://github.com/deepmind/graph_nets.git. Original paper: https://arxiv.org/abs/1806.01261

License:Apache License 2.0