[jira] [Updated] (OPENNLP-1010) NaiveBayes serialized models are not working

["William Colen (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/OPENNLP-1010?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

William Colen updated OPENNLP-1010:
-----------------------------------
    Description: 
The behaviour of a serialized model is different from a model that was never serialized. For example, sometimes using the serialized we get only one outcome, while it was working perfectly during cross validation (when we use it before serializing).
The attached test reproduces it. It compares the outcome array of a model that was never serialized and the same model that was serialized and loaded.

  was:
Behaviour of the serialized model is different of the was never serialized. For example, sometimes we get only one outcome, always.
The attached test reproduces it. It compares the outcome array of the model that was never serialized and one that was loaded from serialized.


> NaiveBayes serialized models are not working
> --------------------------------------------
>
>                 Key: OPENNLP-1010
>                 URL: https://issues.apache.org/jira/browse/OPENNLP-1010
>             Project: OpenNLP
>          Issue Type: Bug
>          Components: Machine Learning
>    Affects Versions: 1.7.2
>            Reporter: William Colen
>         Attachments: NaiveBayesSerializedCorrectnessTest.java
>
>
> The behaviour of a serialized model is different from a model that was never serialized. For example, sometimes using the serialized we get only one outcome, while it was working perfectly during cross validation (when we use it before serializing).
> The attached test reproduces it. It compares the outcome array of a model that was never serialized and the same model that was serialized and loaded.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)