naywhayare changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
sumedh__ has joined #mlpack
sumedh_ has quit [Ping timeout: 240 seconds]
Anand has joined #mlpack
Anand has quit [Ping timeout: 246 seconds]
sumedh__ has quit [Ping timeout: 240 seconds]
Anand has joined #mlpack
< Anand>
Marcus : You saw my mail?
< marcus_zoq>
Anand: Hello, yeah. you are using the correct config?
< Anand>
Yes. You are able to run?
< marcus_zoq>
Anand: Yes
< Anand>
Ok. I must be making some mistake then
< marcus_zoq>
Anand: For me it looks like you are using 'format: [csv]' instead of 'format: [arff, csv]'.
< Anand>
No. I am using the correct format
< Anand>
Anyways, I will see. It seems to be an error at my end. The code is fine. Should I make the merge now?
< marcus_zoq>
Anand: Yeah go ahead!
< Anand>
Ok. In a few minutes
< Anand>
Marcus : Done!
< marcus_zoq>
Anand: Okay great!
< Anand>
I want to take HMM next before bootstrapping.
< Anand>
I will not take more than 2-3 days, I guess and meanwhile you can tell me your opinion on bootstrapping
< marcus_zoq>
Anand: Thanks, sounds great.
< Anand>
Any pointers?
< marcus_zoq>
Anand: Should be pretty straightforward, I think there are only two libraries...
< Anand>
Ok. Mlpack and Matlab, right?
< marcus_zoq>
Anand: Yeah, right!
< Anand>
Do we have the required arrays/files (predictions/probs) for mlpack?
< marcus_zoq>
Anand: I think so, it should be output.csv (hmm viterbi)
< Anand>
Ok. I will see. What about HMM_loglik?
< Anand>
Yes, output.csv contains the predicted state sequence. I need to have a look at how it looks.
< Anand>
Marcus : The output.csv is not getting generated as of now. And yes, I commented the __del__(..) function and gmon.out is getting generated
< marcus_zoq>
Anand: Do you use the version from the gsoc14 branch?
< marcus_zoq>
Anand: In this case you need to rename the RunMethod function.
< marcus_zoq>
Anand: Maybe you should sync the gsoc14 branch with the master branch.
< Anand>
[INFO ] Loading 'datasets/iris_train.csv' as CSV data. Size is 5 x 150. datasets/iris_test.csv:1: parser error : Start tag expected, '<' not found 5.1,3.5,1.4,0.2 ^ [FATAL] Could not load XML file 'datasets/iris_test.csv'!
< marcus_zoq>
Anand: The second file contains the model (xml)
< Anand>
Oh! How can I generate this file?
< marcus_zoq>
Anand: With hmm_train but you can also use the existing model file artificial_2DSignal_hmm.xml.
< Anand>
Ok.
< Anand>
Now it works! The output.csv contains a series of zeroes
< marcus_zoq>
Anand: Yeah, I think we should use some real data.
< Anand>
Real data?
< Anand>
And how does HMM fit to our metrics?
< Anand>
These are state sequences not predicted labels, right?
< marcus_zoq>
Anand: Yeah, right, maybe we should skip the hmm model.
< Anand>
You mean we shoukdn't do this?
< marcus_zoq>
Anand: Yes
< Anand>
Ok, may be for now. But, I will find out if somehow we can use the metrics for HMM.
< Anand>
So, what should I take next?
< marcus_zoq>
Anand: We have forgotten to update the mlpy linear regression method.
Anand_ has joined #mlpack
< Anand_>
mlpy? Ok. I will look into it. It must also have logistic regression then
< marcus_zoq>
Anand: Not necessarily.
< Anand_>
Yeah, it doesn't
< Anand_>
Only linear regression
< Anand_>
I will do it today
< marcus_zoq>
Anand: Btw. I think it would be helpful to store the results of the metrics into the database.
< Anand_>
I thought we are doing that. The benchmark.db?
Anand has quit [Ping timeout: 246 seconds]
< marcus_zoq>
Anand: If we set Log=true we store the timing results into the database.
< Anand_>
Ok. So, we need to run the metrics for all methods once, then.
< marcus_zoq>
Anand: And we should polish the output. Right now we use the Debug output maybe we can print a table?