一流高职院校建设计划项目网站北京和隆优化是干嘛的

import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
import seaborn as sns
import scipy as sp
import sklearn as sk
import time
import gc%matplotlib inline
%config InlineBackend.figure_format = 'retina'import warnings
warnings.filterwarnings("ignore")
plt.rcParams['figure.figsize'] = (10.0, 8.0)
pd.set_option('display.max_columns', 10)
pd.set_option('display.max_rows', None)
pd.set_option('display.min_rows', 14)
pd.set_option('display.max_colwidth',100)
# pd.reset_option('display.max_columns')

import dgl
import torch
import torch.nn as nn
import torch.nn.functional as Ffrom importlib import reload

Using backend: pytorch

导入数据集

import dgl.datadataSet = dgl.data.CoraGraphDataset()

Using backend: pytorchLoading from cache failed, re-processing.
Finished data loading and preprocessing.NumNodes: 2708NumEdges: 10556NumFeats: 1433NumClasses: 7NumTrainingSamples: 140NumValidationSamples: 500NumTestSamples: 1000
Done saving data into cached files.

访问数据集 ndata edata

# 数据集名称
dataSet.name

'cora_v2'

# 查看有哪些图
dataSet.graph

<networkx.classes.digraph.DiGraph at 0x7fdf0458e290>

# 查看label个数
dataSet.num_classes

7

# 一个数据集可能有多个graph 通过下标访问graph
dataSet[0]

Graph(num_nodes=2708, num_edges=10556,ndata_schemes={'train_mask': Scheme(shape=(), dtype=torch.bool), 'val_mask': Scheme(shape=(), dtype=torch.bool), 'test_mask': Scheme(shape=(), dtype=torch.bool), 'label': Scheme(shape=(), dtype=torch.int64), 'feat': Scheme(shape=(1433,), dtype=torch.float32)}edata_schemes={})

节点信息

• 为什么训练集 验证集 测试集的节点数量之和不等于节点总数？

dataSet[0].ndata.keys()

dict_keys(['train_mask', 'val_mask', 'test_mask', 'label', 'feat'])

# 训练集 mask
dataSet[0].ndata['train_mask']

tensor([ True,  True,  True,  ..., False, False, False])

# 训练集包含140个节点
dataSet[0].ndata['train_mask'].sum()

tensor(140)

# 验证集包含500个节点
dataSet[0].ndata['val_mask'].sum()

tensor(500)

# 测试集
dataSet[0].ndata['test_mask']

tensor([False, False, False,  ...,  True,  True,  True])

# 所有节点的分类
dataSet[0].ndata['label']

tensor([3, 4, 4,  ..., 3, 3, 3])

# 所有节点的features
dataSet[0].ndata['feat']

tensor([[0., 0., 0.,  ..., 0., 0., 0.],[0., 0., 0.,  ..., 0., 0., 0.],[0., 0., 0.,  ..., 0., 0., 0.],...,[0., 0., 0.,  ..., 0., 0., 0.],[0., 0., 0.,  ..., 0., 0., 0.],[0., 0., 0.,  ..., 0., 0., 0.]])

边信息

dataSet[0].edata

{}

构造模型并训练

• 见dgl_1_gcn.py

构造dgl graph并查看图信息

# src_ids = [0, 0, 0]
src_ids = torch.tensor([0, 0, 0])
# dest_ids = [1, 2, 3]
dest_ids = torch.tensor([1, 2, 3])

graph = dgl.graph((src_ids, dest_ids), num_nodes = 4)'''
dgl.graph params:data: (src nodes id, dest nodes id) 一一对应
num_nodes: 节点个数构造有向图，边直接由起始节点和终止节点决定，边的id由定义顺序决定
'''

'\ndgl.graph params:\n\ndata: (src nodes id, dest nodes id) 一一对应\nnum_nodes: 节点个数\n\n构造有向图，边直接由起始节点和终止节点决定，边的id由定义顺序决定\n'

# 查看边
graph.edges()

(tensor([0, 0, 0]), tensor([1, 2, 3]))

graph.num_edges()

3

# 给定起始和终止节点查看边id
graph.edge_id(2, 0)

---------------------------------------------------------------------------DGLError                                  Traceback (most recent call last)<ipython-input-23-71bec222622a> in <module>1 # 给定起始和终止节点查看边id
----> 2 graph.edge_id(2, 0)~/anaconda3/envs/DeepLearning/lib/python3.7/site-packages/dgl/heterograph.py in edge_id(self, u, v, force_multi, return_uv, etype)2713         dgl_warning("DGLGraph.edge_id is deprecated. Please use DGLGraph.edge_ids.")2714         return self.edge_ids(u, v, force_multi=force_multi,
-> 2715                              return_uv=return_uv, etype=etype)2716 2717     def edge_ids(self, u, v, force_multi=None, return_uv=False, etype=None):~/anaconda3/envs/DeepLearning/lib/python3.7/site-packages/dgl/heterograph.py in edge_ids(self, u, v, force_multi, return_uv, etype)2835                 raise DGLError("Error: (%d, %d) does not form a valid edge." % (2836                     F.as_scalar(F.gather_row(u, idx)),
-> 2837                     F.as_scalar(F.gather_row(v, idx))))2838             return F.as_scalar(eid) if is_int else eid2839 DGLError: Error: (2, 0) does not form a valid edge.

graph.edge_id(0, 2)

1

# 查看点
graph.nodes()

tensor([0, 1, 2, 3])

graph.num_nodes()

4

# 全图出度之和
graph.num_src_nodes()

4

# 全图入度之和
graph.num_dst_nodes()

4

# 某个点的出度
graph.out_degrees(1)

0

# 某个点的入度
graph.in_degrees(1)

1

# 邻接矩阵
graph.adj()

tensor(indices=tensor([[0, 0, 0],[1, 2, 3]]),values=tensor([1., 1., 1.]),size=(4, 4), nnz=3, layout=torch.sparse_coo)

graph.adjacency_matrix_scipy()

<4x4 sparse matrix of type '<class 'numpy.int64'>'with 3 stored elements in Compressed Sparse Row format>

g, _= dgl.load_graphs('./data/graph.bin', [0])
g = g[0]
g

Graph(num_nodes=4, num_edges=3,ndata_schemes={'feat': Scheme(shape=(3,), dtype=torch.float32)}edata_schemes={'feat': Scheme(shape=(4,), dtype=torch.float32)})

g.srcdata

{'feat': tensor([[ 1.0556,  0.1928,  1.0018],[ 2.2849, -1.3030,  1.3342],[ 1.8428,  0.2911, -0.0281],[-0.3295,  0.4281, -1.9331]])}

g.srcdata.update({'a': torch.tensor([1, 2, 3, 4])})
# g.srcdata['a'] = torch.tensor([1, 2, 3, 4])

g.srcdata

{'feat': tensor([[ 1.0556,  0.1928,  1.0018],[ 2.2849, -1.3030,  1.3342],[ 1.8428,