Python-PyTorchGeometric用于PyTorch的几何深度学习扩展库

[pypi-image]: https://badge.fury.io/py/torch-geometric.svg [pypi-url]: https://pypi.python.org/pypi/torch-geometric [build-image]: https://travis-ci.org/rusty1s/pytorch_geometric.svg?branch=master [build-url]: https://travis-ci.org/rusty1s/pytorch_geometric [docs-image]: https://readthedocs.org/projects/pytorch-geometric/badge/?version=latest [docs-url]: https://pytorch-geometric.readthedocs.io/en/latest/?badge=latest [coverage-image]: https://codecov.io/gh/rusty1s/pytorch_geometric/branch/master/graph/badge.svg [coverage-url]: https://codecov.io/github/rusty1s/pytorch_geometric?branch=master [contributing-image]: https://img.shields.io/badge/contributions-welcome-brightgreen.svg?style=flat [contributing-url]: https://github.com/rusty1s/pytorch_geometric/blob/master/CONTRIBUTING.md <p align="center"> <img width="40%" src="https://raw.githubusercontent.com/rusty1s/pytorch_geometric/master/docs/source/_static/img/pyg_logo_text.svg?sanitize=true" /> </p> -------------------------------------------------------------------------------- [![PyPI Version][pypi-image]][pypi-url] [![Build Status][build-image]][build-url] [![Docs Status][docs-image]][docs-url] [![Code Coverage][coverage-image]][coverage-url] [![Contributing][contributing-image]][contributing-url] **[Documentation](https://pytorch-geometric.readthedocs.io)** | **[Paper](https://arxiv.org/abs/1903.02428)** | **[External Resources](https://pytorch-geometric.readthedocs.io/en/latest/notes/resources.html)** *PyTorch Geometric* (PyG) is a geometric deep learning extension library for [PyTorch](https://pytorch.org/). It consists of various methods for deep learning on graphs and other irregular structures, also known as *[geometric deep learning](http://geometricdeeplearning.com/)*, from a variety of published papers. In addition, it consists of an easy-to-use mini-batch loader for many small and single giant graphs, multi gpu-support, a large number of common benchmark datasets (based on simple interfaces to create your own), and helpful transforms, both for learning on arbitrary graphs as well as on 3D meshes or point clouds. -------------------------------------------------------------------------------- PyTorch Geometric makes implementing Graph Neural Networks a breeze (see [here](https://pytorch-geometric.readthedocs.io/en/latest/notes/create_gnn.html) for the accompanying tutorial). For example, this is all it takes to implement the [edge convolutional layer](https://arxiv.org/abs/1801.07829): ```python import torch from torch.nn import Sequential as Seq, Linear as Lin, ReLU from torch_geometric.nn import MessagePassing class EdgeConv(MessagePassing): def __init__(self, F_in, F_out): super(EdgeConv, self).__init__(aggr='max') # "Max" aggregation. self.mlp = Seq(Lin(2 * F_in, F_out), ReLU(), Lin(F_out, F_out)) def forward(self, x, edge_index): # x has shape [N, F_in] # edge_index has shape [2, E] return self.propagate(edge_index, x=x) # shape [N, F_out] def message(self, x_i, x_j): # x_i has shape [E, F_in] # x_j has shape [E, F_in] edge_features = torch.cat([x_i, x_j - x_i], dim=1) # shape [E, 2 * F_in] return self.mlp(edge_features) # shape [E, F_out] ``` In detail, the following methods are currently implemented: * **[SplineConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.SplineConv)** from Fey *et al.*: [SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernels](https://arxiv.org/abs/1711.08920) (CVPR 2018) * **[GCNConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.GCNConv)** from Kipf and Welling: [Semi-Supervised Classification with Graph Convolutional Networks](https://arxiv.org/abs/1609.02907) (ICLR 2017) * **[ChebConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.ChebConv)** from Defferrard *et al.*: [Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering](https://arxiv.org/abs/1606.09375) (NIPS 2016) * **[NNConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.NNConv)** from Gilmer *et al.*: [Neural Message Passing for Quantum Chemistry](https://arxiv.org/abs/1704.01212) (ICML 2017) * **[ECConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.ECConv)** from Simonovsky and Komodakis: [Edge-Conditioned Convolution on Graphs](https://arxiv.org/abs/1704.02901) (CVPR 2017) * **[GATConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.GATConv)** from Veličković *et al.*: [Graph Attention Networks](https://arxiv.org/abs/1710.10903) (ICLR 2018) * **[SAGEConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.SAGEConv)** from Hamilton *et al.*: [Inductive Representation Learning on Large Graphs](https://arxiv.org/abs/1706.02216) (NIPS 2017) * **[GraphConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.GraphConv)** from, *e.g.*, Morris *et al.*: [Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networks](https://arxiv.org/abs/1810.02244) (AAAI 2019) * **[GatedGraphConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.GatedGraphConv)** from Li *et al.*: [Gated Graph Sequence Neural Networks](https://arxiv.org/abs/1511.05493) (ICLR 2016) * **[GINConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.GINConv)** from Xu *et al.*: [How Powerful are Graph Neural Networks?](https://arxiv.org/abs/1810.00826) (ICLR 2019) * **[ARMAConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.ARMAConv)** from Bianchi *et al.*: [Graph Neural Networks with Convolutional ARMA Filters](https://arxiv.org/abs/1901.01343) (CoRR 2019) * **[SGConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.SGConv)** from Wu *et al.*: [Simplifying Graph Convolutional Networks](https://arxiv.org/abs/1902.07153) (CoRR 2019) * **[APPNP](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.APPNP)** from Klicpera *et al.*: [Predict then Propagate: Graph Neural Networks meet Personalized PageRank](https://arxiv.org/abs/1810.05997) (ICLR 2019) * **[AGNNConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.AGNNConv)** from Thekumparampil *et al.*: [Attention-based Graph Neural Network for Semi-Supervised Learning](https://arxiv.org/abs/1803.03735) (CoRR 2017) * **[RGCNConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.RGCNConv)** from Schlichtkrull *et al.*: [Modeling Relational Data with Graph Convolutional Networks](https://arxiv.org/abs/1703.06103) (ESWC 2018) * **[SignedConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.SignedConv)** from Derr *et al.*: [Signed Graph Convolutional Network](https://arxiv.org/abs/1808.06354) (ICDM 2018) * **[DNAConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.DNAConv)** from Fey: [Just Jump: Dynamic Neighborhood Aggregation in Graph Neural Networks](https://arxiv.org/abs/1904.04849) (ICLR-W 2019) * **[EdgeConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.EdgeConv)** from Wang *et al.*: [Dynamic Graph CNN for Learning on Point Clouds](https://arxiv.org/abs/1801.07829) (CoRR, 2018) * **[PointConv](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.conv.PointConv)** (including **[Iterative Farthest Point Sampling](https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.pool.fps)**, dynamic graph generation based on