Probabilistic classification of the Fermi-LAT 4FGL-DR3 catalog sources

Multi-class classification of Fermi LAT 4FGL-DR3 sources into six or nine classes using random forest (RF) and neural network (NN) algorithms. Probabilistic catalogs are in files: 4FGL-DR3_6class_GMM_nmin100_prob_cat.csv<br> 4FGL-DR3_6class_RF_nmin100_prob_cat.csv<br> 4FGL-DR3_9class_GMM_nmin15_prob_cat.csv Files containing definition of the six or nine groups of physical classes of the 4FGL-DR3 catalog and summary of the expected numbers of associated and unassociated sources in each of the six or nine groups: 4FGL-DR3_6class_GMM_nmin100_summary.csv<br> 4FGL-DR3_6class_RF_nmin100_summary.csv<br> 4FGL-DR3_9class_GMM_nmin15_summary.csv The identifies in file names, GMM_nmin100, RF_nmin100, GMM_nmin15, refer to the method how the groups of physical classes were determined: (1) using the Gaussian mixture model (GMM) or random forest (RF), (2) requiring at least 100 associated sources in a group (nmin100) or at least 15 sources in a group (nmin15). A detailed description can be found at: http://arxiv.org/abs/2301.07412