A CNN pruning approach using constrained binary particle swarm optimization with a reduced search space for image classification

Jihene Tmamna, Emna Ben Ayed, Rahma Fourati, Amir Hussain, Mounir Ben Ayed

Abstract: Highlights•CNN filter pruning is regarded as a constrained single-objective optimization problem.•A search space reduction strategy is performed before the evolutionary process.•A modified BPSO initialization method based on the L1<math><msub is="true"><mrow is="true"><mi is="true">L</mi></mrow><mrow is="true"><mn is="true">1</mn></mrow></msub></math> norm criterion is proposed.•During the search, BPSO-FPruner avoids redundant evaluations, reducing complexity.•Validation on different pre-trained models with carbon footprint evaluation is handled.