"Parallel Processing Extension"
Hello all,
I have a dataset with nearly 100 attributes/columns. I am using this dataset in order to train the Neural Net operator for prediction.
However the run time is too too too long.
Found the parallel processing extension, but I do not know how to use it for my case.
Can i find any tutorial or material which helps me to use this extension to speed up the run time of the training phase?
Thanks in advance.
Lakshmi
I have a dataset with nearly 100 attributes/columns. I am using this dataset in order to train the Neural Net operator for prediction.
However the run time is too too too long.
Found the parallel processing extension, but I do not know how to use it for my case.
Can i find any tutorial or material which helps me to use this extension to speed up the run time of the training phase?
Thanks in advance.
Lakshmi
Find more posts tagged with
Sort by:
1 - 2 of
21
thanks for the alternatives.
I am trying to reduce the number of features.
thanks again for the wonderful support.
I am trying to reduce the number of features.
thanks again for the wonderful support.
with the Parallel Extension you can't parallelize the internals of an operator, that means that the neural net will need the same amount of time as before. You can only parallelize the execution of multiple operators, e.g. if you use you neural net in an X-Validation, you could exchange the validation operator with the X-Validation (Parallel).
So to speed up the training of the learner itself, you have threeoptions (or a combination of them):
1. reduce the number of features by using means of feature selection (Forward Selection or Weight by XXX combined with Select by Weights)
2. reduce the number examples with the Sample operator
3. use another learner. For data with a lot of features e.g. the SVM is often a good choice, since its runtime is linear in the number of features. Plus it delivers far more interpretable models than the Neural Net
Best, Marius