bkloppenborg / simtoi

The SImulation and Modeling Tool for Optical Interferometry
GNU General Public License v3.0
7 stars 6 forks source link

Doesn't find a starspot solution #135

Closed therob762 closed 9 years ago

therob762 commented 9 years ago

I have attached the .json file of the 256x256 model I was running on the attached data file. In short, this is a single starspot on the surface of a rapid rotator where multinest is run to find the polar radius, the power-law limb darkening coefficient, and the four starspot parameters. This is the output when I first start the model:

"X Error: BadWindow (invalid Window parameter) 3
  Major opcode: 20 (X_GetProperty)
  Resource id:  0x7c0004b
Data set information for: 
 /home/parksj/research/data/inter/sig_per/sig_Per.oifits
N Vis: 0
N V2 : 216
N T3 : 272
N UV : 1028
Average JD: 2455828.96234
X Error: BadWindow (invalid Window parameter) 3
  Major opcode: 20 (X_GetProperty)
  Resource id:  0x7c00051
 *****************************************************
 MultiNest v3.9
 Copyright Farhan Feroz & Mike Hobson
 Release Oct 2014

 no. of live points =  200
 dimensionality =    6
 *****************************************************
 Starting MultiNest
 generating live points"

I let the code run for over 12 hours and the only output that was generated is when I exit SIMTOI. This is the output upon exit:

"Exit signal received
Could not read multinestsummary.txt file.
Best-fit parameters and exported values will be invalid.
terminate called after throwing an instance of 'std::range_error'
  what():  The new value for R_pole is less than the set minimum.
Aborted (core dumped)"

Multinest seemed to be running the entire time, however just prior to exiting it didn't seem any closer to the solution than when it began.

Okay, so I can't attach the .json file, but here is the text:

{
    "_file_info" : "Save file from the SImulation and Modeling Tool for Optical Interferometry (SIMTOI) version 1.1.0 commit 069c52f. See https://github.com/bkloppenborg/simtoi for information about this format",
    "area_height" : 256,
    "area_scale" : 0.050000000000000003,
    "area_width" : 256,
    "model_0" : 
    {
        "base_data" : 
        {
            "T_eff_pole" : [ 5000, 2000, 1000000, false, 100, 0 ],
            "g_pole" : [ 1, 0.01, 10000, false, 10, 2 ],
            "inclination" : [ 0, -180, 180, false, 1, 2 ],
            "n_side_power" : [ 4, 1, 10, false, 1, 0 ],
            "omega_rot" : [ 0.5, 0, 1, false, 0.10000000000000001, 2 ],
            "position_angle" : [ 0, 0, 360, false, 0.10000000000000001, 2 ],
            "r_pole" : [ 3, 2.5, 3.5, true, 1, 4 ],
            "von_zeipel_beta" : [ 0.5, 0.01, 1, false, 0.10000000000000001, 2 ],
            "z_axis_rotation" : [ 0, 0, 360, false, 1, 2 ],
            "z_axis_rotational_period" : [ 0, 0, 100, false, 1, 4 ]
        },
        "base_id" : "roche_rotator",
        "feature_0_data" : 
        {
            "delta_T" : [ -500, -2000, 0, true, 100, 0 ],
            "phi" : [ 0, 0, 180, true, 0.10000000000000001, 2 ],
            "radius" : [ 1, 0, 10, true, 1, 2 ],
            "theta" : [ 0, 0, 180, true, 0.10000000000000001, 2 ]
        },
        "feature_0_id" : "uniform_spot",
        "position_data" : 
        {
            "E" : [ 0, -1, 1, false, 0.10000000000000001, 2 ],
            "N" : [ 0, -1, 1, false, 0.10000000000000001, 2 ]
        },
        "position_id" : "xy",
        "shader_data" : 
        {
            "alpha" : [ 0.25, 0.10000000000000001, 0.5, true, 0.089999999999999997, 1 ]
        },
        "shader_id" : "ldl_power_law"
    }
}
fabienbaron commented 9 years ago

To make things clearer, Multinest doesn't seem to be able to generate all the live points, so it never gets to the minimization part.

bkloppenborg commented 9 years ago

After @therob762 and I investigated this issue more, we found the v2err field in the data file contained -1 values. This resulted in SIMTOI sending NAN values to MultiNest which, in turn, caused the solution not to progress. In fbbc25effa5abf0466cbfa58b2749e63c197df13 I pulled in an updated version of ccoifits which tests for -1 in the error field and (silently) sets flags in the error column. This is now fixed in the develop branch. Be sure to do a git pull and git submodule update to get the revised version of ccoifits