[PIPE2D-510] Evaluate the 2D RBF interpolation results Created: 29/Jan/20  Updated: 05/Jan/21  Resolved: 11/Mar/20

Status: Done
Project: DRP 2-D Pipeline
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Task Priority: Normal
Reporter: Takuji Yamashita Assignee: Takuji Yamashita
Resolution: Done Votes: 0
Labels: flux-calibration, model-templates
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: PNG File p_rms_alpha.png     PNG File p_rms_logg.png     PNG File p_rms_numspectra.png     PNG File p_rms_teff.png     PNG File p_rms_z.png     PNG File processingtime_numspectra.png     PNG File processingtime.png     PNG File s_rms_alpha.png     PNG File s_rms_logg.png     PNG File s_rms_numspectra.png     PNG File s_rms_teff.png     PNG File s_rms_z.png    
Issue Links:
Blocks
blocks PIPE2D-365 Interpolate between model templates Done
Story Points: 2
Epic Link: flux calibration
Sprint: 2DDRP-2020 A, 2DDRP-2021 A
Reviewers: hassan

 Description   

Evaluate the accuracy of the parameter interpolation of the AMBRE model spectra with the 2D RBF interpolation. 



 Comments   
Comment by Takuji Yamashita [ 29/Jan/20 ]

For calculating the accuracies, interpolated spectra were generated at an original parameter grid and were compared with an original model template at its grid. Numbers of interpolated spectra are 2590 and 2125 for parallel plan models and spherical models, respectively.
 
Processing time is 8-9 sec per an interpolation on average. The time was improved from the 4D interpolation because used model templates were reduced from ~100-300 to ~30. 
 
Processing time (sec)

  mean min max median stddev
p-model 8.636729 8.308514 9.230231 8.643057 0.103503
s-model 8.295978 7.713312 10.262160 8.278359 0.218529

 
The RMS of residuals is on average less than 1% but it becomes worse (>1%) for the blue side of spherical models. 
 
RMS (%)

  mean min max median stddev
p-model all 0.049745 0.001470 1.144177 0.014773 0.103231
p-model blue 0.082114 0.001504 2.027684 0.023470 0.178944
p-model nir 0.019759 0.000869 0.377890 0.006437 0.037863
p-model red 0.025319 0.000738 0.473988 0.008091 0.049938
s-model all 0.769161 0.012876 21.199334 0.300931 1.726265
s-model blue 1.356829 0.024040 40.370744 0.509499 3.146068
s-model nir 0.260185 0.002558 4.715903 0.129480 0.436969
s-model red 0.340520 0.004064 6.427464 0.165988 0.596282

 

Comment by Takuji Yamashita [ 29/Jan/20 ]

The 2D interpolation has been done along two parameters of Teff and Log(g) at a fixed Z and [alpha/Fe]. The code is faster than one of 4D but the accuracies do not satisfy the required one (<1%).

Comment by Takuji Yamashita [ 14/Feb/20 ]

The evaluation has been done. The script for 2D RBF interpolation was uploaded in GitHub. 

Generated at Sat Feb 10 15:54:15 JST 2024 using Jira 8.3.4#803005-sha1:1f96e09b3c60279a408a2ae47be3c745f571388b.