矩阵补全IGMC 学习笔记

目录

Inductive Graph-based Matrix Completion (IGMC) 模型

igmc推理示例：

---

Inductive Graph-based Matrix Completion (IGMC) 模型

原版代码：

IGMC/models.py at master · muhanzhang/IGMC · GitHub

GNN推理示例

torch_geometric版本：torch_geometric-2.5.3

原版报错，edge_type找不到，通过删除参数修正的：

import torch
import torch.nn.functional as F
from torch.nn import Linear
from torch_geometric.nn import GCNConv, global_add_pool
from torch_geometric.utils import dropout_adj
from torch_geometric.data import Data, DataLoader
class GNN(torch.nn.Module):
    # a base GNN class, GCN message passing + sum_pooling
    def __init__(self, dataset, gconv=GCNConv, latent_dim=[32, 32, 32, 1],
                 regression=False, adj_dropout=0.2, force_undirected=False):
        super(GNN, self).__init__()
        self.regression = regression
        self.adj_dropout = adj_dropout
        self.force_undirected = force_undirected
        self.convs = torch.nn.ModuleList()
        self.convs.append(gconv(dataset.num_features, latent_dim[0]))
        for i in range(0, len(latent_dim)-1):
            self.convs.append(gconv(latent_dim[i], latent_dim[i+1]))
        self.lin1 = Linear(sum(latent_dim), 128)
        if self.regression:
            self.lin2 = Linear(128, 1)
        else:
            self.lin2 = Linear(128, dataset.num_classes)

    def reset_parameters(self):
        for conv in self.convs:
            conv.reset_parameters()
        self.lin1.reset_parameters()
        self.lin2.reset_parameters()

    def forward(self, data):
        x, edge_index, batch = data.x, data.edge_index, data.batch
        if self.adj_dropout > 0:
            edge_index, _ = dropout_adj(
                edge_index, p=self.adj_dropout,
                force_undirected=self.force_undirected, num_nodes=len(x),
                training=self.training
            )
        concat_states = []
        for conv in self.convs:
            x = torch.tanh(conv(x, edge_index))
            concat_states.append(x)
        concat_states = torch.cat(concat_states, 1)
        x = global_add_pool(concat_states, batch)
        x = F.relu(self.lin1(x))
        x = F.dropout(x, p=0.5, training=self.training)
        x = self.lin2(x)
        if self.regression:
            return x[:, 0]
        else:
            return F.log_softmax(x, dim=-1)

    def __repr__(self):
        return self.__class__.__name__

# 创建一个简单的数据类，用于模拟数据集属性
class SimpleDataset:
    num_features = 2
    num_classes = 2

# 创建一个简单的图数据集
edge_index = torch.tensor([[0, 1, 2, 3], [1, 0, 3, 2]], dtype=torch.long)
x = torch.tensor([[1, 0], [0, 1], [1, 0], [0, 1]], dtype=torch.float)
batch = torch.tensor([0, 0, 1, 1], dtype=torch.long)

# 使用 Data 类构建图数据
data = Data(x=x, edge_index=edge_index, batch=batch)

# 构建 DataLoader
loader = DataLoader([data], batch_size=2, shuffle=False)

dataset = SimpleDataset()

# 实例化模型
model = GNN(dataset)

# 模型推理
model.eval()
for data in loader:
    out = model(data)
    print(out)

igmc推理示例：

import torch.nn as nn
import torch.nn.functional as F
from torch.nn import Linear, Conv1d
from torch_geometric.nn import GCNConv, RGCNConv, global_sort_pool, global_add_pool
from torch_geometric.utils import dropout_adj
from util_functions import *
import pdb
import time
from torch_geometric.data import Data, DataLoader
class GNN(torch.nn.Module):
    # a base GNN class, GCN message passing + sum_pooling
    def __init__(self, dataset, gconv=GCNConv, latent_dim=[32, 32, 32, 1],
                 regression=False, adj_dropout=0.2, force_undirected=False):
        super(GNN, self).__init__()
        self.regression = regression
        self.adj_dropout = adj_dropout
        self.force_undirected = force_undirected
        self.convs = torch.nn.ModuleList()
        self.convs.append(gconv(dataset.num_features, latent_dim[0]))
        for i in range(0, len(latent_dim)-1):
            self.convs.append(gconv(latent_dim[i], latent_dim[i+1]))
        self.lin1 = Linear(sum(latent_dim), 128)
        if self.regression:
            self.lin2 = Linear(128, 1)
        else:
            self.lin2 = Linear(128, dataset.num_classes)

    def reset_parameters(self):
        for conv in self.convs:
            conv.reset_parameters()
        self.lin1.reset_parameters()
        self.lin2.reset_parameters()

    def forward(self, data):
        x, edge_index, batch = data.x, data.edge_index, data.batch
        if self.adj_dropout > 0:
            # edge_index, edge_type = dropout_adj(
            #     edge_index, edge_type, p=self.adj_dropout,
            #     force_undirected=self.force_undirected, num_nodes=len(x),
            #     training=self.training
            # )
            edge_index, edge_type = dropout_adj(edge_index, p=self.adj_dropout, force_undirected=self.force_undirected, num_nodes=len(x), training=self.training)
        concat_states = []
        for conv in self.convs:
            x = torch.tanh(conv(x, edge_index))
            concat_states.append(x)
        concat_states = torch.cat(concat_states, 1)
        x = global_add_pool(concat_states, batch)
        x = F.relu(self.lin1(x))
        x = F.dropout(x, p=0.5, training=self.training)
        x = self.lin2(x)
        if self.regression:
            return x[:, 0]
        else:
            return F.log_softmax(x, dim=-1)

    def __repr__(self):
        return self.__class__.__name__
class IGMC(GNN):
    # The GNN model of Inductive Graph-based Matrix Completion.
    # Use RGCN convolution + center-nodes readout.
    def __init__(self, dataset, gconv=RGCNConv, latent_dim=[32, 32, 32, 32],
                 num_relations=5, num_bases=2, regression=False, adj_dropout=0.2,
                 force_undirected=False, side_features=False, n_side_features=0,
                 multiply_by=1):
        super(IGMC, self).__init__(
            dataset, GCNConv, latent_dim, regression, adj_dropout, force_undirected
        )
        self.multiply_by = multiply_by
        self.convs = torch.nn.ModuleList()
        self.convs.append(gconv(dataset.num_features, latent_dim[0], num_relations, num_bases))
        for i in range(0, len(latent_dim)-1):
            self.convs.append(gconv(latent_dim[i], latent_dim[i+1], num_relations, num_bases))
        self.lin1 = Linear(2*sum(latent_dim), 128)
        self.side_features = side_features
        if side_features:
            self.lin1 = Linear(2*sum(latent_dim)+n_side_features, 128)

    def forward(self, data):
        start = time.time()
        x, edge_index, edge_type, batch = data.x, data.edge_index, data.edge_type, data.batch
        if self.adj_dropout > 0:
            edge_index, edge_type = dropout_adj(
                edge_index, edge_type, p=self.adj_dropout,
                force_undirected=self.force_undirected, num_nodes=len(x),
                training=self.training
            )
        concat_states = []
        for conv in self.convs:
            x = torch.tanh(conv(x, edge_index, edge_type))
            concat_states.append(x)
        concat_states = torch.cat(concat_states, 1)

        users = data.x[:, 0] == 1
        items = data.x[:, 1] == 1
        x = torch.cat([concat_states[users], concat_states[items]], 1)
        if self.side_features:
            x = torch.cat([x, data.u_feature, data.v_feature], 1)

        x = F.relu(self.lin1(x))
        x = F.dropout(x, p=0.5, training=self.training)
        x = self.lin2(x)
        if self.regression:
            return x[:, 0] * self.multiply_by
        else:
            return F.log_softmax(x, dim=-1)

class SimpleDataset:
    num_features = 2
    num_classes = 2

# 创建一个简单的图数据集
edge_index = torch.tensor([[0, 1, 2, 3], [1, 0, 3, 2]], dtype=torch.long)
edge_type = torch.tensor([0, 1, 2, 3], dtype=torch.long)
x = torch.tensor([[1, 0], [0, 1], [1, 0], [0, 1]], dtype=torch.float)
batch = torch.tensor([0, 0, 1, 1], dtype=torch.long)

# 使用 Data 类构建图数据
data = Data(x=x, edge_index=edge_index,edge_type=edge_type, batch=batch)

# 构建 DataLoader
loader = DataLoader([data], batch_size=2, shuffle=False)

dataset = SimpleDataset()

# 实例化模型
model = IGMC(dataset)

# 模型推理
model.eval()
for data in loader:
    out = model(data)
    print(out)