Why I put a pair of same graph and it gives me 0.5 of similarity?

[Lost-little-dinosaur]

Hi~
I ran the code, it is logical and clear, I like it.
But I when I try to let it compute my own gragh data, I am confuse...
I have change it's default training time and let it train for 1000 times, it's lose perfrom like this:

It is very good, isn't it? The lose is bellow 0.01. But when I try to input two same gragh data (like this)

and hope it can give me 1 of similarity, but I get a similarity about 0.5
Then I try some other graph data that have different size, they all give me a similarity between 0.39-0.5

Why is that? I can show you how do I get the similarity:

Is there any problem? Can you pleace help me? Thank you in advance.

[Lost-little-dinosaur]

And I also want to know the meaning of the label

I know that the label array means the feature of each node...But how can I change the feature in to more complex one such as string?
In fact, I am trying to use this project to compare two ast trees that come from two code project, so my label's value will be the source code of the two project... I will calulate the relationship between by calulate the similaraty of each string...

Can you help me, pleace?

[Lost-little-dinosaur]

I have already gave this project a star....
Can you help me, please?

[lower01]

I have already gave this project a star.... Can you help me, please?

Hello brother, like you, I constructed the graph data myself and then output the similarity based on prediction. Even for the same data, it is around 0.5, which I can’t understand. Do you understand the problem now?

[lower01]

Hello brother, like you, I constructed the graph data myself and then output the similarity based on prediction. Even for the same data, it is around 0.5, which I can’t understand. Do you understand the problem now?

[zcyyyy24]

Hi~ I ran the code, it is logical and clear, I like it. But I when I try to let it compute my own gragh data, I am confuse... I have change it's default training time and let it train for 1000 times, it's lose perfrom like this: It is very good, isn't it? The lose is bellow 0.01. But when I try to input two same gragh data (like this) and hope it can give me 1 of similarity, but I get a similarity about 0.5 Then I try some other graph data that have different size, they all give me a similarity between 0.39-0.5

Why is that? I can show you how do I get the similarity: Is there any problem? Can you pleace help me? Thank you in advance.

同学你好！我想问一下你是怎么计算两张图之间的相似度？我也是想把节点特征增加几个，不知道这样SimGNN能不能计算相似度？

[zcyyyy24]

Hello brother, like you, I constructed the graph data myself and then output the similarity based on prediction. Even for the same data, it is around 0.5, which I can’t understand. Do you understand the problem now?

同学你好！请问你是怎么计算两张图之间的相似度？如果想增加节点特征该怎么修改代码呢？