Contents

Input file templateΒΆ

This is what the input file of parameters looks like. Everything is in here and it should be used as a template for your data. Note the blank space between the variable and its value. This input file is case insensitive and after each run a copy of the input file is saved along with the results
#Input parameters for MLZ (Don't delete them)
#Space between ":" and "value" is important
#
#            BLANK SPACE
#                 |                  
#                 | 
#                 |
# FILENAMES       V
TrainFile        : SDSS_MGS.train     # Name of training file
TestFile         : SDSS_MGS.test      # Name of test file
FinalFileName    : SDSS_MGS           # Root of Final name
#------------------------------------------------------
#     PATHS
#------------------------------------------------------
Path_Train       : test/  # Path to train file
Path_Test        : test/  # Path to test file
Path_Output      : test/  # Path to results
#------------------------------------------------------
#     COLUMNS AND ATTRIBUTES
#------------------------------------------------------
#in columns names the column for error must be named same a the column with an e in front
Columns          : zs,u,g,r,i,z,u-g,g-r,r-i,i-z,eu,eg,er,ei,ez,eu-g,eg-r,er-i,ei-z   # Columns of training file
Att              : u,g,r,i,z,u-g,g-r,r-i,i-z                  # Attributes to use, try a subset!
Columns_Test     : zs,u,g,r,i,z,u-g,g-r,r-i,i-z,eu,eg,er,ei,ez,eu-g,eg-r,er-i,ei-z     # Columns for test, no need have a KetAtt column
KeyAtt           : zs                           #name of  the column to be predicted
#------------------------------------------------------
#     CODE STEPS
#------------------------------------------------------
CheckOnly        : no                # Check everything with small samples to test the codes
PredictionMode   : TPZ               # TPZ/SOM/TPZ_C
PredictionClass  : Reg               # Reg/Class
#------------------------------------------------------
#   REDSHIFT RELATED (OR FOR CLASSIFICATION)
#------------------------------------------------------
MinZ             : 0.001             # Min redshift (class label)
MaxZ             : 0.3               # Max redshift (class label)
NzBins           : 80                # Number of bins in z range (2 for Classification), Final output is double this value
#------------------------------------------------------
#   COMMON FOR TPZ/SOM
#------------------------------------------------------
NRandom          : 2                 # Number of random realizations (1 original data)
NTrees           : 4                 # Number of trees or maps!  (total number is NRandom x NTrees)
Natt             : 3                 # Number of attributes m* for TPZ or subsample for SOM
OobError         : no                # Out-of-bag error?
VarImportance    : no                # Perform variable importance ranking?
#------------------------------------------------------
#   TPZ
#------------------------------------------------------
MinLeaf          : 5                 # Minimum number for terminal leaf
ImpurityIndex    : entropy           # options (entropy/gini/classE) for Classification
#------------------------------------------------------
#   SOM
#------------------------------------------------------
Topology         : hex               # grid,hex,sphere
Periodic         : yes               # periodic? yes/no
Ntop             : 15                # map size (in the case fo sphere, npix=12*ntop*ntop , ntop: must be power of 2)
Iterations       : 200               # number of iterations
SomType          : online            # online/batch
AlphaStart       : 0.9               # Alpha start
AlphaEnd         : 0.5               # Alpha end
ImportanceFile   : none              # path to importance file, write 'none' is not used
#------------------------------------------------------
#   Get PDFs
#------------------------------------------------------
SigmaFactor      : 3.                # Gaussian smoothing with kernel Sigma1*Resolution
RmsFactor        : 0.02              # initial rms, can be modified using oob data
WriteFits        : no               # yes, PDF file in fits format, no: numpy format
MultipleFiles    : no                # Write a PDFs file per core instead of merging them (faster)
#------------------------------------------------------
#   SPARSE REPRESENTATION
#------------------------------------------------------
SparseRep        : no                # Sparse representation, it creates a fits file
SparseDims       : 200,50,2          # Nmu, Nsigma, Nv (To create dictionary)
NumberCoef       : 32001             # Integers to represent the coefficients
NumberBases      : 20                # Number of fixed bases to use
OriginalPdfFile  : yes                # Write the original PDFs also?
#------------------------------------------------------

Contents