Ogee spillways with converging training walls are applied to lower the hazard of accidental flooding in locations with limited construction operations due to their unique structure. Hence, this type of structure is proposed as an emergency spillway. The present study aimed at experimental and machine learning-based modeling of the submerged discharge capacity of the converging ogee spillway. Two experimental models of Germi-Chay dam spillway were utilized: one model having a curve axis which was made in 1:50 scale and the other with a straight axis in 1:75 scale. Using visual observation, it was found that the total upstream head, the submergence degree, the ogee-crest geometries and the convergence angle of training walls are the crucial factors which alter the submerged discharge capacity of the converging ogee spillway. Furthermore, two machine-learning techniques (e.g. artificial neural networks and gene expression programming) were applied for modeling the submerged discharge capacity applying experimental data. These models were compared with four well-known traditional relationships with respect to their basic theoretical concept. The obtained results indicated that the length ratio () had the most effective role in estimating the submerged discharge capacity.