Documentation:HDF5 schema

From ALPS
Jump to: navigation, search

Contents

The HDF5 schema

Observables within ALPS are written according to a specific hierarchical naming convention which is listed below.

Parameters

All the parameters specified in the input parameter files are stored in:

/parameters/PARAMETERNAME  : parameter value

Monte Carlo Data

Each observable with a given OBSERVABLENAME has the following quantities stored in the hdf5-outputfile:

/simulation/results/OBSERVABLENAME/count  : number of measurements
/simulation/results/OBSERVABLENAME/mean/value  : mean value of the observable
/simulation/results/OBSERVABLENAME/mean/error  : error of the observable
/simulation/results/OBSERVABLENAME/mean/error_convergence  : error convergence of the observable
/simulation/results/OBSERVABLENAME/variance/value  : variance of the observable
/simulation/results/OBSERVABLENAME/tau/value  : autocorrelation time of the observable
/simulation/results/OBSERVABLENAME/jacknife/data  : jacknife data of the observable
/simulation/results/OBSERVABLENAME/timeseries/data  : timeseries of the mean of the observable
/simulation/results/OBSERVABLENAME/timeseries/data2  : timeseries of the mean of the observable squared
/simulation/results/OBSERVABLENAME/timeseries/partialbin  : mean of the observable in the partialbin
/simulation/results/OBSERVABLENAME/timeseries/partialbin2  : mean of the observable squared in the partialbin
/simulation/results/OBSERVABLENAME/timeseries/logbinning  : mean of the observable in the logbin
/simulation/results/OBSERVABLENAME/timeseries/logbinning2  : mean of the observable squared in the logbin
/simulation/results/OBSERVABLENAME/timeseries/logbinning_last  : mean of the observable in the last logbin
/simulation/results/OBSERVABLENAME/timeseries/logbinning_count  : depth of the logbinning tree

To be deprecated? :

/simulation/results/OBSERVABLENAME/sum  : mean of the observable
/simulation/results/OBSERVABLENAME/sum2  : mean of the observable squared


In the non-aggregated hdf5-files (with the file name parm.taski.out.runj.h5) the path to the data stored is of the form: /simulation/realizations/NUMBER/clones/MPIPROCESSNUMBER/results/OBSERVABLENAME/... , where the system realization (e.g. disorder realization) is enumerated with NUMBER and the clone with the mpi process number MPIPROCESSNUMBER .

ED or DMRG Data

For observables obtained from an ED or DMRG simulation the data is stored at the following paths within the hdf5-output file:

/spectrum/energies  : the energy eigenvalues
/spectrum/sectors/NUMBER/energies  : the energy values for a specific sector
/spectrum/sectors/NUMBER/quantumnumbers/QUANTUMNUMBERNAME  : the quantum number value of QUANTUMNUMBERNAME for a specific sector
/spectrum/results/OBSERVABLENAME/mean/value  : expectation value of the observable
/spectrum/results/OBSERVABLENAME/labels  : labels for the x-coordinate of the observable
/spectrum/iteration/NUMBER/results/OBSERVABLENAME/mean/value  : expectation value of the observable for a specific iteration (DMRG only)
/spectrum/iteration/NUMBER/results/OBSERVABLENAME/labels  : labels for the x-coordinate of the observable


Some results that do not depend on the individual states targeted for is instead at /simulation/results with a array value containing the results for the individual iterations.

Data Attributes

  • jacknife/data has the attribute:
binningtype  : the binning scheme, schemes are: linear or logarithmic
  • timeseries/data and timeseries/data2 have the attributes:
maxbinnum  : the maximum number of bins kept
minbinsize  : minimal number of measurements per bin
binsize  : number of measurements per bin
binningtype  : the binning scheme, schemes are: linear or logarithmic
discard (deprecated)  : bins discarded
  • /timeseries/partialbin and /timeseries/partialbin2 have the attribute:
count  : binsize of the partialbin
  • timeseries/logbinning, timeseries/logbinning2, timeseries/logbinning_lastbin and timeseries/logbinning_counts have the attribute:
binningtype  : the binning scheme, schemes are: linear or logarithmic
  • A HistogramObservable has the attributes:
size  : number of entries in the histogram
min  : minimal x-value recordable in the histogram
max  : maximal x-value recordable in the histogram
stepsize  : the width of the interval dx in the histogram
  • A SignedObservable has the attribute:
sign  : observable name of the sign observable, default: Sign
  • Observables can have the attributes:
cannotrebin  : whether or not rebinning is possible, i.e. whether or not a nonlinear transformation has been performed
changed (deprecated)  : whether or not the data has been transformed
nonlinearoperations (deprecated)  : whether or not a nonlinear operation has been performed on the data
Personal tools