This paper claimed that the introduction of so-called graph

Hi <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="

Where is the implementation of `Graph Size Normalization`? about benchmarking-gnns HOT 1 CLOSED

graphdeeplearning commented on August 21, 2024

Where is the implementation of `Graph Size Normalization`?

from benchmarking-gnns.

Comments (1)

vijaydwivedi75 commented on August 21, 2024

Hi @Dinxin,

It is implemented in an earlier version of the repo in every <data>.py file in the data/ directory.

For instance for ZINC:

benchmarking-gnns/data/molecules.py

Lines 165 to 167 in a17fdf1

    
           tab_sizes_n = [ graphs[i].number_of_nodes() for i in range(len(graphs))] 
        
           tab_snorm_n = [ torch.FloatTensor(size,1).fill_(1./float(size)) for size in tab_sizes_n ] 
        
           snorm_n = torch.cat(tab_snorm_n).sqrt()

The values thus calculated are used at a GNN layer as:

benchmarking-gnns/layers/gcn_layer.py

Lines 58 to 59 in a17fdf1

    
           if self.graph_norm: 
        
               h = h * snorm_n # normalize activation w.r.t. graph size

from benchmarking-gnns.

Where is the implementation of `Graph Size Normalization`? about benchmarking-gnns HOT 1 CLOSED

Comments (1)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	tab_sizes_n = [ graphs[i].number_of_nodes() for i in range(len(graphs))]
	tab_snorm_n = [ torch.FloatTensor(size,1).fill_(1./float(size)) for size in tab_sizes_n ]
	snorm_n = torch.cat(tab_snorm_n).sqrt()

	if self.graph_norm:
	h = h * snorm_n # normalize activation w.r.t. graph size