PREDNASAJUCI / LECTURER : Julius Koza NAZOV / TITLE : Semi-empirical spectral modeling in hindsight ABSTRAKT / ABSTRACT : Semi-empirical spectral modeling is a fundamental technique in diagnostic of astrophysical plasma. In essence, it can be characterized by the set of (i) observations (data) used to constrain the model, (ii) physical approximations (formulas) employed, and (iii) free parameters ("model") to be determined. Data is inevitably limited in spectral, spatial, and temporal domains, modified by instrumental characteristics, and burdened by omnipresent noise. Physical approximations are complete only to the level of making a problem tractable. An adopted complexity of model is justified by a need to be computationally affordable. A balance in the three-point system "data-formulas-model" is quantified by a multi-variate merit function. Its global minimum represents the sought solution, which is often identified with physical reality. The presentation aims to show an application of the technique on solar flare loops, long off-limb spicule, small-scale chromospheric jet, and quiescent filament. Although in all these cases a good fit of data, i.e. a balance in the system "data-formulas-model", was reached, the final model parameters may be considered as suspicious and even unrealistic. Only in particular case of cool flare loops the final model is considered to be adequate and justified because substantially different data and physical approximations yielded very similar results. Conclusively, the presentation aims to demonstrate that a data fit does not imply ultimate physical adequacy and realism of model parameters but rather a balance in the "data-formulas-model" system.