评估每个超边的重要性的代码

[13121283123]

对您的这篇论文非常感兴趣，想问您一下，您在定义layers1的时候 只用到了BatchedGraphSAGEDynamicRangeMean1 这个类，其他剩下的类您注释了BatchedGraphSAGEDynamicMean1，BatchedGraphSAGEMean1，BatchedGraphSAGEMean1Temporal，BatchedGAT_cat1 请问这几个类有用吗，还有想请问您一下 ， for i in range(int(N/p)):
idx_start = max(0, i-t)
idx_end = min(i+t+1, int(N/p))
tmp_x = x[:,idx_startp:idx_endp,] dis = NearestConvolution.cos_dis(tmp_x) 这个是计算特征之间的相似性 if i==0: tk = min(dis.shape[2], self.kn)

print(tk)

        _, idx = torch.topk(dis, tk, dim=2) 是包含前K个近邻的邻居集
        k_nearest = torch.stack([torch.stack([tmp_x[j, idx[j, i]] for i in range(p*(idx_end-idx_start))], dim=0) for j in range(b)], dim=0) #(b, x*p, kn, d)
        #print(k_nearest)
        k_nearest_list.append(k_nearest[:,p*(i-idx_start):p*(i-idx_start+1),])
    k_nearest = torch.cat(k_nearest_list, dim=1) #(b,N, kn, d)
    x_neib = k_nearest[:,:,1:,].contiguous()  我们将除节点 v i 以外的超边中的所有节点特征进行平均操作,作为超边的特征.
    x_neib = x_neib.mean(dim=2)
    h_k = torch.cat((self.W_x(x), self.W_neib(x_neib)), 2)

    h_k = F.normalize(h_k, dim=2, p=2)
    h_k = F.relu(h_k)
    #print(h_k.shape)
    if self.use_bn:
        #self.bn = nn.BatchNorm1d(h_k.size(1))
        h_k = self.bn(h_k.permute(0,2,1).contiguous())
        #print(h_k.shape)
        h_k = h_k.permute(0, 2, 1)
        #print(h_k.shape)

    return h_k

请问评估每个超边的重要性的代码 在哪里啊 谢谢您的回复

[daodaofr]

其他几个类是我们尝试过的不同的特征传播的方法，应该效果比现在用的稍微差点。 这里就是简单进行了节点的平均作为超边特征，我们发现attention对结果影响较小，这里就直接平均化了。

[13121283123]

非常感谢您的回复，还想请教您一个问题， 当调用self.layers1的时候会跳到BatchedGraphSAGEDynamicRangeMean1类的forward方法，但是我看到这个forward方法中没有用到传入进来的adj邻阶矩阵，这个怎么解释啊， 大佬， 谢谢 class BatchedGraphSAGEDynamicRangeMean1(nn.Module): def init(self, infeat, outfeat, use_bn=True, mean=False, add_self=False): super(BatchedGraphSAGEDynamicRangeMean1, self).init() self.add_self = add_self self.use_bn = use_bn self.mean = mean self.aggregator = True self.W_x = nn.Linear(infeat, outfeat, bias=True) nn.init.xavieruniform(self.W_x.weight, gain=nn.init.calculate_gain('relu')) self.W_neib = nn.Linear(infeat, outfeat, bias=True) nn.init.xavieruniform(self.W_neib.weight, gain=nn.init.calculate_gain('relu')) if self.use_bn: self.bn = nn.BatchNorm1d(2*outfeat)

self.bn = nn.BatchNorm1d(16)

self.kn = 3

def forward(self, x, adj, p, t): b = x.size()[0]
N = x.size()[1]
k_nearest_list = []
tk = self.kn #tk=3 for i in range(int(N/p)):
idx_start = max(0, i-t)
idx_end = min(i+t+1, int(N/p))
tmp_x = x[:,idx_startp:idx_endp,] dis = NearestConvolution.cos_dis(tmp_x) if i==0: tk = min(dis.shape[2], self.kn)

print(tk)

    _, idx = torch.topk(dis, tk, dim=2)

    k_nearest = torch.stack([torch.stack([tmp_x[j, idx[j, i]] for i in range(p*(idx_end-idx_start))], dim=0) for j in range(b)], dim=0) 
    k_nearest_list.append(k_nearest[:,p*(i-idx_start):p*(i-idx_start+1),])
k_nearest = torch.cat(k_nearest_list, dim=1) #(b,N, kn, d)
x_neib = k_nearest[:,:,1:,].contiguous()
#x_neib = x_neib.view(x.size(0), x.size(1), -1, x_neib.size(2))
x_neib = x_neib.mean(dim=2)
#print(k_nearest.shape)
#x_cmp = x - k_nearest[:,:,0]
#print(torch.sum(x_cmp)) 
h_k = torch.cat((self.W_x(x), self.W_neib(x_neib)), 2)
h_k = F.normalize(h_k, dim=2, p=2)
h_k = F.relu(h_k)
if self.use_bn:
    #self.bn = nn.BatchNorm1d(h_k.size(1))
    h_k = self.bn(h_k.permute(0,2,1).contiguous())
    #print(h_k.shape)
    h_k = h_k.permute(0, 2, 1)
    #print(h_k.shape)
return h_k

[daodaofr]

这里是用p，t 这两个参数来控制他的邻居的选取，就没用邻接矩阵

[13121283123]

好的好的 谢谢 ------------------ 原始邮件 ------------------ 发件人: "daodaofr/hypergraph_reid" @.>; 发送时间: 2021年4月8日(星期四) 晚上8:17 @.>; 抄送: "Changxia @.**@.>; 主题: Re: [daodaofr/hypergraph_reid] 评估每个超边的重要性的代码 (#8)

这里是用p，t 这两个参数来控制他的邻居的选取，就没用邻接矩阵

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.