Multi video summarization using query based deep optimization algorithm

Ansari, Shaharyar Alam; Zafar, Aasim

doi:10.1007/s13042-023-01852-3

Cited by 3 publications

(1 citation statement)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For instance, designing water supply systems, high speed cold tandem mill, and developing land use management strategies all involve complex MaOPs. While existing algorithms can efficiently handle 2D or 3D MOPs, their performance severely deteriorates for MaOPs with more than three objectives [3]. Developing methods to effectively solve MaOPs has thus become an critical research area [4].…”

Section: Introductionmentioning

confidence: 99%

Improving Many-Objective Optimization via An Adaptive Penalty Scheme and Jaya Operator in the MJaya/D Algorithm

Zhang,

Yang,

Xiong

et al. 2024

Preprint

View full text Add to dashboard Cite

Achieving an effective balance between solution convergence and diversity remains a key challenge when solving multi-objective optimization problems. However, the performance of existing algorithms continues to degrade on many-objective optimization problems. Additionally, the genetic operators commonly utilized in many-objective evolutionary algorithms require extensive tuning of control parameters, without which solution quality suffers. To address these limitations, several new techniques are proposed. First, an adaptive penalty mechanism that leverages population and weight vector distribution information to assign specialized penalty factors to each sub-problem is developed by us. Second, the Jaya optimization algorithm is extended by us to improve its efficacy in many-objective spaces. Finally, Levy mutation is incorporated to help escape local optima and enhance population diversity. The proposed MJaya/D algorithm is assessed on standard DTLZ benchmark problems to evaluate the performance improvements stemming from the Jaya operator and adaptive penalty scheme. Experiments demonstrate that MJaya/D obtains superior Pareto front approximations compared to current state-of-the-art techniques for several many-objective optimization formulations.

show abstract

Section: Introductionmentioning

confidence: 99%