“…DCNN design complexity has been reduced by metaheuristics (Khishe and Mosavi, 2020a , b ; Mosavi et al, 2017 ; Qiao et al, 2021 ), which develop an architecture without the assistance of a human designer (Rastogi and Choudhary, 2019 ). Numerous metaheuristics have been effectively studied and applied, including genetic programming (GP) (Suganuma et al, 2017 ), particle swarm optimization (PSO) (Mosavi and Khishe, 2017 ; Wang et al, 2018 ; J. Wu et al, 2021a , b ), Position-transitional particle swarm optimization (Luo et al, 2020 ), sine–cosine algorithm (SCA) (Wang et al, 2020 ; C. Wu et al, 2021a , b ), chimp optimization algorithm (ChOA) (Hu et al, 2021b ), salp swarm algorithm (Khishe and Mohammadi, 2019 ), and genetic algorithms (GAs) (Liu et al, 2022a , b ). However, learning from large data sets is impracticable due to the high cost of computing and the time of the learning process (Yuan et al, 2020 ).…”