Search alternatives:
driven optimization » process optimization (Expand Search)
design optimization » bayesian optimization (Expand Search)
binary data » primary data (Expand Search), dietary data (Expand Search)
Showing 1 - 20 results of 52 for search '(( binary _ driven optimization algorithm ) OR ( binary data design optimization algorithm ))', query time: 0.47s Refine Results
  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15

    The Pseudo-Code of the IRBMO Algorithm. by Chenyi Zhu (9383370)

    Published 2025
    “…In order to comprehensively verify the performance of IRBMO, this paper designs a series of experiments to compare it with nine mainstream binary optimization algorithms. …”
  16. 16
  17. 17

    IRBMO vs. meta-heuristic algorithms boxplot. by Chenyi Zhu (9383370)

    Published 2025
    “…In order to comprehensively verify the performance of IRBMO, this paper designs a series of experiments to compare it with nine mainstream binary optimization algorithms. …”
  18. 18

    IRBMO vs. feature selection algorithm boxplot. by Chenyi Zhu (9383370)

    Published 2025
    “…In order to comprehensively verify the performance of IRBMO, this paper designs a series of experiments to compare it with nine mainstream binary optimization algorithms. …”
  19. 19

    Proposed Algorithm. by Hend Bayoumi (22693738)

    Published 2025
    “…Hence, an Energy-Harvesting Reinforcement Learning-based Offloading Decision Algorithm (EHRL) is proposed. EHRL integrates Reinforcement Learning (RL) with Deep Neural Networks (DNNs) to dynamically optimize binary offloading decisions, which in turn obviates the requirement for manually labeled training data and thus avoids the need for solving complex optimization problems repeatedly. …”
  20. 20

    Comparisons between ADAM and NADAM optimizers. by Hend Bayoumi (22693738)

    Published 2025
    “…Hence, an Energy-Harvesting Reinforcement Learning-based Offloading Decision Algorithm (EHRL) is proposed. EHRL integrates Reinforcement Learning (RL) with Deep Neural Networks (DNNs) to dynamically optimize binary offloading decisions, which in turn obviates the requirement for manually labeled training data and thus avoids the need for solving complex optimization problems repeatedly. …”