Open Access

Table 1

Summary of the predictive performance for our three different cINN setups and the three different point estimation methods.

cINN

Cube selection Point estimator Measure Parameter Single LoS Neighbour LoS Single LoS (7 wavelengths)
MAP NRMSE log(n/m−3) 0.0707 0.0608 0.0981
log(Tdust/K) 0.0300 0.0260 0.0458
   
|ērel| (%) log(n/m−3)
log(Tdust/K)
 
All cubes MeanShift NRMSE log(n/m−3) 0.0640 0.0555 0.0855
log(Tdust/K) 0.0251 0.0221 0.0372
   
|ērel| (%) log(n/m−3)
log(Tdust/K)
 
MNPCP NRMSE log(n/m−3) 0.0620 0 0536 0.0838
log(Tdust/K) 0.0245 0.0213 0.0359
   
|ērel| (%) log(n/m−3)
log(Tdust/K)

NRMSE log(n/m−3) 0.0510 0.0441 0.0917
MAP log(Tdust/K) 0.0633 0.0550 0.1135
   
|ērel| (%) log(n/m−3)
log(Tdust/K)
 
ISRF-only MeanShift NRMSE log(n/m−3) 0.0480 0.0409 0.0804
log(Tdust/K) 0.0522 0.0455 0.0906
   
|ērel| (%) log(n/m−3)
log(Tdust/K)
 
MNPCP NRMSE log(n/m−3) 0.0456 0.0395 0.0772
log(Tdust/K) 0.0506 0.0439 0.0859
   
|ērel| (%) log(n/m−3)
log(Tdust/K)

MAP NRMSE log(n/m−3) 0.0859 0.0739 0.1041
log(Tdust/K) 0.0387 0.0336 0.0538
   
|ērel| (%) log(n/m−3)
log(Tdust/K)
 
ISRF + star MeanShift NRMSE log(n/m−3) 0.0767 0.0670 0.0903
log(Tdust/K) 0.0327 0.0289 0.0441
   
|ērel| (%) log(n/m−3)
log(Tdust/K)
 
MNPCP NRMSE log(n/m−3) 0.0748 0 0648 0.0899
log(Tdust/K) 0.0320 0.0279 0.0431
   
|ērel| (%) log(n/m−3)
log(Tdust/K)

Notes. Listed are respectively the NRMSE and the median absolute relative error |ērel| (along with the 25% and 75% quantiles) for the dust density and temperature evaluated across all pixels of the set of test cubes indicated in the first column.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.