thieu1995/mealpy v1.0.4

Changed category of Sparrow Search Algorithm (SpaSA) from Fake to Swarm-based group:

• Added the: OriginalSpaSA
  • This version is taken from the original paper, very weak algorithm
• BaseSpaSA: My changed version
  • Changed equations
  • Changed flows and operators
  • This version become the BEST algorithm

Added Jaya Algorithm to Swarm-based group:

• OriginalJA: The original version from original paper
• BaseJA: My version of original JA for better running time.
  • Remove all third loop in algorithm
  • Change the second random variable r2 to Gaussian instead of uniform
• LJA: The original version of: Levy-flight Jaya Algorithm (LJA)
  • Paper: An improved Jaya optimization algorithm with Levy flight
  • Link: https://doi.org/10.1016/j.eswa.2020.113902
  • Notes:
    • This version I still remove all third loop in algorithm
    • The beta value of Levy-flight equal to 1.8 as the best value in the paper.

DE, its state-of-the-art variants.

• DESAP: including DESAP-Abs and DESAP-Rel
  • The main ideas is identified the population size without user-defined. Proposed equation:
    • Initial ps_init = 10*n (n: is the problem size, number of dimensions)
    • DESAP-Abs: ps = round(ps_init + N (0, 1)), (N: is Gaussian value)
    • DESAP-Rel: ps = round(ps_init + U (-0.5, 0.5)), (U: is uniform random function)

Added Battle Royale Optimization Algorithm to Fake-algorithm

• OriginalBRO:
  • The paper is very different than the author's matlab code. Even the algorithm's flow is wrong with index i, j.
  • I tested the results is very slow convergence, even with small #dimensions. I guess that is why he cloned the crossover process of Genetic Algorithm to his algorithm in the code (but not even mention it in the paper) to get the results in the paper. Don't know what to say about this.
• BaseBRO:
  • First, I removed all third loop in the algorithm for faster computation.
  • Second, Re-defined the algorithm's flow and algorithm's ideas

Added Fruit-fly Optimization Algorithm and its variants to Swarm-based group:

• OriginalFOA:
  • This algorithm is the weakest algorithm in MHAs. It can't run with complicated objective function.
• BaseFOA:
  • I changed the fitness function (smell function) by taking the distance each 2 adjacent dimensions --> Number of variables reduce from N to N-1
  • Update the position if only it find the better fitness value.

• WFOA:

  • The original version of Whale Fruit-fly Optimization Algorithm (WFOA)
  • Paper: Boosted Hunting-based Fruit Fly Optimization and Advances in Real-world Problems

  • From my point of view, this algorithm is almost the same as Whale, only different in calculate fitness function. So it is not surprise that It outperforms BaseFOA

    https://www.sciencedirect.com/science/article/abs/pii/S0957417420307545 https://sci-hub.se/10.1016/j.eswa.2020.113976 https://sci-hub.se/10.1016/j.eij.2020.08.003 https://sci-hub.se/10.1016/j.eswa.2020.113902 https://www.x-mol.com/paper/1239433029684543488

Update root.py

• Added improved_ms() function based on mutation and search mechanism - current better than levy-flight technique

• Add FBIO examples with large-scale benchmark functions