Hi Pavel,

I'm not sure, I've never tried this. I do not know this error. Do both F and G have the same number of measurements?

Could you please try to create a small test example that reproduces the error so we can have a look at it?

Emanuel

On Jul 18, 2012, at 11:38 AM, august@fzu.cz wrote:

Hi Emanuel,

thank you for quick answer. I changed my measurement objects from 'SimpleRealVectorObservable' to 'RealVectorObservable' and now it works. To some extent.. I can calculate the covariance matrix of a directly measured quantity like Im G. But when I want to calculate the covariance matrix for the selfenergy (calculated from the equation of motion as F/G) and I run the code on more then one CPU, it fails again. This time, the error is:

Bin number: 87 86 Bin size: 128 128 terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injectorstd::runtime_error

...

'

what(): both observables need same number of measurements and bins

which sounds a bit surprising since I call it as ReSigma.covariance(ReSigma). Any suggestions?

Thank you again!

Pavel

...

Hi Pavel,

are you trying to compute covariances of 'SimpleRealVectorObservable' objects or of collected objects? The 'RealVectorObservable' objects should have all the information needed for the covariances.

In the next version, hybridization3, we write out the covariance matrices to the hdf5 by default. In that way they can be used directly in Maxent. This is probably what you need.

Best regards, Emanuel

On Jul 18, 2012, at 10:31 AM, august@fzu.cz wrote:

...

Hi all,

I would like to calculate the covarince matrix of MC data from CT-HYB. But when I run the method "covarinace" of the SimpleObservableEvaluator class, it fails on "no binning information available for calculation of covariances". Is there some way to fix this and set the number of bins? I am using ALPS 2.0.2-r5790. Thank you,

Pavel

Dear Vipin,

It seems that the HDF5 installation on your machine was removed, or that you copied an ALPS executable to a new machine that did not have HDF5 installed. The best will be to make sure that HDF5 is still installed on your machine, reinstall it if necessary, and then build ALPS again.

Matthias

On Jul 19, 2012, at 3:24 AM, Vipin Varma wrote:

Dear ALPS users,

When I try to convert the parameter file to an xml file (parameter2xml), I get an error - I guess from the linker - complaining of a missing shared library file libhdf5.so.6. This happens in a couple of my alps-2 installations, all of which I had used sometime back and don't remember changing anything since then. Trying to export to a local environment variable should not help because the file itself is missing.

Can someone please mention if there is a minimal fix for this, without having to install everything again?

Thanks and regards, Vipin

Hi Emanuel,

Do you need a standalone code that will fail on this error? I tried to modify the

alps-2.0.2-r5790-src-with-boost/alps/test/alea/simpleobseval.C

file, but I don't know how to run it in such a way that measurements are done on different CPUs and then evaluated together (on one CPU it always works even in my case).

Anyway, the problem seems to be that ReG and ImG do does not have the same number of bins even though they are measured and collected together.

Structure of the simplest failing code is the following:

data types used: std::valarray<double> re_G_omega_meas; std::valarray<double> im_G_omega_meas;

measurements << RealVectorObservable("re_G_omega"); measurements << RealVectorObservable("im_G_omega"); . . Measure ReG and ImG.. re_G_omega_meas[i] = G.real(); im_G_omega_meas[i] = G.imag(); . . collect measurements: measurements.get <RealVectorObservable> ("re_G_omega") << re_G_omega_meas; measurements.get <RealVectorObservable> ("im_G_omega") << im_G_omega_meas; . . evaluation: alps::RealVectorObsevaluator reG, imG; reG = get_measurements()["re_G_omega"]; imG = get_measurements()["im_G_omega"];

reG.covariance(imG); And here, it fails with the error mesage I sent in my last mail ("both observables need same number of measurements and bins").

Is this enough? If not, I will try to produce some real code.

Pavel

Hi Pavel,

I'm not sure, I've never tried this. I do not know this error. Do both F and G have the same number of measurements?

Could you please try to create a small test example that reproduces the error so we can have a look at it?

Emanuel

On Jul 18, 2012, at 11:38 AM, august@fzu.cz wrote:

...

Hi Emanuel,

thank you for quick answer. I changed my measurement objects from 'SimpleRealVectorObservable' to 'RealVectorObservable' and now it works. To some extent.. I can calculate the covariance matrix of a directly measured quantity like Im G. But when I want to calculate the covariance matrix for the selfenergy (calculated from the equation of motion as F/G) and I run the code on more then one CPU, it fails again. This time, the error is:

Bin number: 87 86 Bin size: 128 128 terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injectorstd::runtime_error

...

'

what(): both observables need same number of measurements and bins

which sounds a bit surprising since I call it as ReSigma.covariance(ReSigma). Any suggestions?

Thank you again!

Pavel

...

Hi Pavel,

are you trying to compute covariances of 'SimpleRealVectorObservable' objects or of collected objects? The 'RealVectorObservable' objects should have all the information needed for the covariances.

In the next version, hybridization3, we write out the covariance matrices to the hdf5 by default. In that way they can be used directly in Maxent. This is probably what you need.

Best regards, Emanuel

On Jul 18, 2012, at 10:31 AM, august@fzu.cz wrote:

...

Hi all,

I would like to calculate the covarince matrix of MC data from CT-HYB. But when I run the method "covarinace" of the SimpleObservableEvaluator class, it fails on "no binning information available for calculation of covariances". Is there some way to fix this and set the number of bins? I am using ALPS 2.0.2-r5790. Thank you,

Pavel

Hi,

finally, I found that there is a technical problem on our cluster causing significant slowdown of writing to the hard drive. This is most likely the origin of my problems so you can forget about it. Sorry for spamming the mail list.

Best regards,

Pavel

Hi once again,

it seems that the source of my problem is simple. Probably, the HybridizationSimItime::get_result() method is called before all the MC runs are over. Therefore, during the evaluation process, some data is still being written to the measurement objects and it causes inconsistency. You can check this by trying something like

reG = get_measurements()["re_G_omega"]; cout << reG.count() << endl; . . do something.. . . reG = get_measurements()["re_G_omega"]; cout << reG.count() << endl;

This way, I get two different results. Is there some easy way how to fix this?

Pavel

...

Hi Emanuel,

Do you need a standalone code that will fail on this error? I tried to modify the

alps-2.0.2-r5790-src-with-boost/alps/test/alea/simpleobseval.C

file, but I don't know how to run it in such a way that measurements are done on different CPUs and then evaluated together (on one CPU it always works even in my case).

Anyway, the problem seems to be that ReG and ImG do does not have the same number of bins even though they are measured and collected together.

Structure of the simplest failing code is the following:

data types used: std::valarray<double> re_G_omega_meas; std::valarray<double> im_G_omega_meas;

measurements << RealVectorObservable("re_G_omega"); measurements << RealVectorObservable("im_G_omega"); . . Measure ReG and ImG.. re_G_omega_meas[i] = G.real(); im_G_omega_meas[i] = G.imag(); . . collect measurements: measurements.get <RealVectorObservable> ("re_G_omega") << re_G_omega_meas; measurements.get <RealVectorObservable> ("im_G_omega") << im_G_omega_meas; . . evaluation: alps::RealVectorObsevaluator reG, imG; reG = get_measurements()["re_G_omega"]; imG = get_measurements()["im_G_omega"];

reG.covariance(imG); And here, it fails with the error mesage I sent in my last mail ("both observables need same number of measurements and bins").

Is this enough? If not, I will try to produce some real code.

Pavel

...

Hi Pavel,

I'm not sure, I've never tried this. I do not know this error. Do both F and G have the same number of measurements?

Could you please try to create a small test example that reproduces the error so we can have a look at it?

Emanuel

On Jul 18, 2012, at 11:38 AM, august@fzu.cz wrote:

...

Hi Emanuel,

thank you for quick answer. I changed my measurement objects from 'SimpleRealVectorObservable' to 'RealVectorObservable' and now it works. To some extent.. I can calculate the covariance matrix of a directly measured quantity like Im G. But when I want to calculate the covariance matrix for the selfenergy (calculated from the equation of motion as F/G) and I run the code on more then one CPU, it fails again. This time, the error is:

Bin number: 87 86 Bin size: 128 128 terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injectorstd::runtime_error

...

'

what(): both observables need same number of measurements and bins

which sounds a bit surprising since I call it as ReSigma.covariance(ReSigma). Any suggestions?

Thank you again!

Pavel

...

Hi Pavel,

are you trying to compute covariances of 'SimpleRealVectorObservable' objects or of collected objects? The 'RealVectorObservable' objects should have all the information needed for the covariances.

In the next version, hybridization3, we write out the covariance matrices to the hdf5 by default. In that way they can be used directly in Maxent. This is probably what you need.

Best regards, Emanuel

On Jul 18, 2012, at 10:31 AM, august@fzu.cz wrote:

...

Hi all,

I would like to calculate the covarince matrix of MC data from CT-HYB. But when I run the method "covarinace" of the SimpleObservableEvaluator class, it fails on "no binning information available for calculation of covariances". Is there some way to fix this and set the number of bins? I am using ALPS 2.0.2-r5790. Thank you,

Pavel