Metadata-Version: 2.1
Name: fastGAT
Version: 0.0.9
Summary: A Fast Graph ATtention networks with ALSH. 
Home-page: UNKNOWN
Author: LiangZi@NEU
Author-email: 2273067585@qq.com
License: UNKNOWN
Description: # Fast Graph ATtention neural networks
        ## What's that?
        Fast Graph ATtention neural networks(fastGAT) is a new algorithm based on GAT(graph attention networks). We designed our algorithm with AngleLSH(Angle based Local Sensity Hashing), for a faster running speed and a less memory usage. 
        In a word, you can use our algorithm instead of GAT in everywhere with more great effect.
        ## How to install it?
        Just use:
        ```
        pip install fastGAT
        ```
        
        ## How to use it?
        ### A demo for fast run.
        you can just use:
        ```python
        import fastGAT.fastGAT as ft
        
        proc=ft.exe()
        proc.tra_val()
        proc.test()
        ```
        for simple demo, which will running our method in CORA dataset.
        
        ### customize your own hyperparameters.
        use
        ```python
        from fasGAT.model import fastGAT as mft
        
        blablabla..
        
        ```
        ### customize your own dataset
        just try the numpy format matrix , which for example： 
        ```
        adj: the adj matrix
        features: feature vector combination of all nodes
        labels: label of all nodes
        ```
        
        ### MORE INFORMATION
        For more infomation, you might go to here:[fastGAT](https://github.com/liangzid/FastGraphATtention) 
        ## Does it well?
        You can enjoy some images follow for the detail of it.
        
        
        ## concat me.
        you can send any problem and BUG using the issue in github.  Or call me with: 2273067585@qq.com.
        
        
        
        
        
        
        
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Description-Content-Type: text/markdown
