Table 18 Runtime distribution of the proposed framework key components on the CelebA and LFW datasets.

From: Reinforcement learning-driven feature selection enhanced by an evolutionary approach tuning for criminal suspect identification

Step

Dataset

CelebA

LFW

Runtime (s)

% of Total

Runtime (s)

% of Total

CNN feature extraction

1964

62

1449

55

Off-policy PPO FS

792

25

685

26

DE hyperparameter tuning

412

13

501

19