GitHub user shahidki31 opened a pull request:

    https://github.com/apache/spark/pull/21627

    [SPARK-24484][MLLIB]Power Iteration Clustering is giving incorrect 
clustering results when there are mutiple leading eigen values.

    ## What changes were proposed in this pull request?
    
![image](https://user-images.githubusercontent.com/23054875/41823325-e83e1d34-781b-11e8-8c34-fc6e7a042f3f.png)
    
    
![image](https://user-images.githubusercontent.com/23054875/41823367-733c9ba4-781c-11e8-8da2-b26460c2af63.png)
    
![image](https://user-images.githubusercontent.com/23054875/41823409-179dd910-781d-11e8-8d8c-9865156fad15.png)
    
    **Method to determine if the top eigen values has same magnitude but 
opposite signs**
    
    
![image](https://user-images.githubusercontent.com/23054875/41822941-f8b13d4c-7814-11e8-8091-54c02721c1c5.png)
    
![image](https://user-images.githubusercontent.com/23054875/41822982-b80a6fc4-7815-11e8-9129-ed96a14f037f.png)
    
![image](https://user-images.githubusercontent.com/23054875/41823022-5b69e906-7816-11e8-847a-8fa5f0b6200e.png)
     
    
![image](https://user-images.githubusercontent.com/23054875/41823087-54311398-7817-11e8-90bf-e1be2bbff323.png)
    
![image](https://user-images.githubusercontent.com/23054875/41823121-e0b78324-7817-11e8-9596-379bd2e518af.png)
    
![image](https://user-images.githubusercontent.com/23054875/41823151-965319d2-7818-11e8-8b91-10f6276ace62.png)
    
![image](https://user-images.githubusercontent.com/23054875/41823182-75cdbad6-7819-11e8-912f-23c66a8359de.png)
    
![image](https://user-images.githubusercontent.com/23054875/41823221-1ca77a36-781a-11e8-9a40-48bd165797cc.png)
    
![image](https://user-images.githubusercontent.com/23054875/41823272-f6962b2a-781a-11e8-9978-1b2dc0dc8b2c.png)
    
![image](https://user-images.githubusercontent.com/23054875/41823303-75b296f0-781b-11e8-8501-6133b04769c8.png)
    
    
    **So, we need to check if the reileigh coefficient at the convergence is 
lesser than the norm of the estimated eigen vector before normalizing**
    
    
    
    
    (Please fill in changes proposed in this fix)
    Added a UT
    
    Please review http://spark.apache.org/contributing.html before opening a 
pull request.


You can merge this pull request into a Git repository by running:

    $ git pull https://github.com/shahidki31/spark picConvergence

Alternatively you can review and apply these changes as the patch at:

    https://github.com/apache/spark/pull/21627.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

    This closes #21627
    
----
commit 53fd9da9209c57ef6e90bba85fa9083f49a8619d
Author: Shahid <shahidki31@...>
Date:   2018-06-24T19:30:43Z

    Power iteration failed to converge

----


---

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to