[
https://issues.apache.org/jira/browse/MAHOUT-968?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13203008#comment-13203008
]
Dirk Weißenborn commented on MAHOUT-968:
----------------------------------------
You need to feed the label and the images file to the MnistPreparer in
Mahout-Examples with size 44000 should work and chunknumber 440. then try
training the network with the directory where the chunks are written. then run
the training just greedy first so use "-nf" for no finetuning as option for the
training. structure should be 784,500,1000 (these are the layers of the
network). if you want to run it with hadoop use -mapreduce. everything else is
ok with the defaults. after training greedily, the model will be saved. you can
take this model to finetune later (set output path to the model, it will be
materialized). use then -ng for nogreedy training. this first part of the
training took me on an 8 core machine about 26 hours!! I have a trained model
if you would like to have that :)
> Classifier based on restricted boltzmann machines
> -------------------------------------------------
>
> Key: MAHOUT-968
> URL: https://issues.apache.org/jira/browse/MAHOUT-968
> Project: Mahout
> Issue Type: New Feature
> Components: Classification
> Affects Versions: 0.7
> Reporter: Dirk Weißenborn
> Labels: classification, mnist
> Fix For: 0.7
>
> Attachments: MAHOUT-968.patch, MAHOUT-968.patch
>
> Original Estimate: 336h
> Remaining Estimate: 336h
>
> This is a proposal for a new classifier based on restricted boltzmann
> machines. The development of this feature follows the paper on "Deep
> Boltzmann Machines" (DBM) [1] from 2009. The proposed model (DBM) got an
> error rate of 0.95% on the mnist dataset [2], which is really good. Main
> parts of the implementation should also be applicable to other scenarios than
> classification where restricted boltzmann machines are used (ref. MAHOUT-375).
> I am working on this feature right now, and the results are promising. The
> only problem with the training algorithm is, that it is still mostly
> sequential (if training batches are small, what they should be), which makes
> Map/Reduce until now, not really beneficial. However, since the algorithm
> itself is fast (for a training algorithm), training can be done on a single
> machine in managable time.
> Testing of the algorithm is currently done on the mnist dataset itself to
> reproduce results of [1]. As soon as results indicate, that everything is
> working fine, I will upload the patch.
> [1] http://www.cs.toronto.edu/~hinton/absps/dbm.pdf
> [2] http://yann.lecun.com/exdb/mnist/
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
