The Educational Competition Optimizer (ECO) is a newly proposed human-based metaheuristic algorithm. It derives from the phenomenon of educational competition in society with good performance. However, the basic ECO is constrained by its limited exploitation and exploration abilities when tackling complex optimization problems and exhibits the drawbacks of premature convergence and diminished population diversity. To this end, this paper proposes an enhanced educational competition optimizer, named EDECO, by incorporating estimation of distribution algorithm and replacing some of the best individual(s) using a dynamic fitness distance balancing strategy. On the one hand, the estimation of distribution algorithm enhances the global exploration ability and improves the population quality by establishing a probabilistic model based on the dominant individuals provided by EDECO, which solves the problem that the algorithm is unable to search the neighborhood of the optimal solution. On the other hand, the dynamic fitness distance balancing strategy increases the convergence speed of the algorithm and balances the exploitation and exploration through an adaptive mechanism. Finally, this paper conducts experiments on the proposed EDECO algorithm with 29 CEC 2017 benchmark functions and compares EDECO with four basic algorithms as well as four advanced improved algorithms. The results show that EDECO indeed achieves significant improvements compared to the basic ECO and other compared algorithms, and performs noticeably better than its competitors. Next, this study applies EDECO to 10 engineering constrained optimization problems, and the experimental results show the significant superiority of EDECO in solving real engineering optimization problems. These findings further support the effectiveness and usefulness of our proposed algorithm in solving complex engineering optimization challenges.

Beluga whale optimization (BWO) is a swarm-based metaheuristic algorithm inspired by the group behavior of beluga whales. BWO suffers from drawbacks such as an insufficient exploration capability and the tendency to fall into local optima. To address these shortcomings, this paper proposes augmented multi-strategy beluga optimization (AMBWO). The adaptive population learning strategy is proposed to improve the global exploration capability of BWO. The introduction of the roulette equilibrium selection strategy allows BWO to have more reference points to choose among during the exploitation phase, which enhances the flexibility of the algorithm. In addition, the adaptive avoidance strategy improves the algorithm’s ability to escape from local optima and enriches the population quality. In order to validate the performance of the proposed AMBWO, extensive evaluation comparisons with other state-of-the-art improved algorithms were conducted on the CEC2017 and CEC2022 test sets. Statistical tests, convergence analysis, and stability analysis show that the AMBWO exhibits a superior overall performance. Finally, the applicability and superiority of the AMBWO was further verified by several engineering optimization problems.

Currently, deepfake detection has garnered widespread attention as a key defense mechanism against the misuse of deepfake technology. However, existing deepfake detection networks still face challenges such as insufficient robustness, limited generalization capabilities, and a single feature extraction domain (e.g., using only spatial domain features) when confronted with evolving algorithms or diverse datasets, which severely limits their application capabilities. To address these issues, this study proposes a deepfake detection network named EFIMD-Net, which enhances performance by strengthening feature interaction and integrating spatial and frequency domain features. The proposed network integrates a Cross-feature Interaction Enhancement module (CFIE) based on cosine similarity, which achieves adaptive interaction between spatial domain features (RGB stream) and frequency domain features (SRM, Spatial Rich Model stream) through a channel attention mechanism, effectively fusing macro-semantic information with high-frequency artifact information. Additionally, an Enhanced Multi-scale Feature Fusion (EMFF) module is proposed, which effectively integrates multi-scale feature information from various layers of the network through adaptive feature enhancement and reorganization techniques. Experimental results show that compared to the baseline network Xception, EFIMD-Net achieves comparable or even better Area Under the Curve (AUC) on multiple datasets. Ablation experiments also validate the effectiveness of the proposed modules. Furthermore, compared to the baseline traditional two-stream network Locate and Verify, EFIMD-Net significantly improves forgery detection performance, with a 9-percentage-point increase in Area Under the Curve on the CelebDF-v1 dataset and a 7-percentage-point increase on the CelebDF-v2 dataset. These results fully demonstrate the effectiveness and generalization of EFIMD-Net in forgery detection. Potential limitations regarding real-time processing efficiency are acknowledged.

Numerous existing methods demonstrate impressive performance in brightening low-illumination images but fail in detail enhancement and color correction. To tackle these challenges, this paper proposes a dual-branch network including three main parts: color space transformation, a color correction network (CC-Net), and a light-boosting network (LB-Net). Specifically, we first transfer the input into the CIELAB color space to extract luminosity and color components. Afterward, we employ LB-Net to effectively explore multiscale features via a carefully designed large–small-scale structure, which can adaptively adjust the brightness of the input images. And we use CC-Net, a U-shaped network, to generate noise-free images with vivid color. Additionally, an efficient feature interaction module is introduced for the interaction of the two branches’ information. Extensive experiments on low-light image enhancement public benchmarks demonstrate that our method outperforms state-of-the-art methods in restoring the quality of low-light images. Furthermore, experiments further indicate that our method significantly enhances performance in object detection under low-light conditions.