An Adaptive Lion Swarm Optimization Algorithm Incorporating Tent Chaotic Search and Information Entropy

Miaomiao Liu, Yuying Zhang, Jingfeng Guo, Jing Chen, Zhigang LiuInternational Journal of Computational Intelligence Systems 2023 · DOI

TL;DR

Researchers have developed ACIE_LSO, an enhanced version of the Lion Swarm Optimization algorithm that integrates information entropy and tent chaotic mapping. By dynamically adjusting how "lion cubs" explore their environment and using chaotic search to escape local traps, the algorithm achieves significantly faster convergence and higher accuracy across complex optimization tasks, including neural network training for house price prediction.

What Problem Does This Paper Address?

Swarm intelligence algorithms—like those inspired by ants, bees, or lions—are powerful tools for finding the "best" solution in a massive search space. However, the standard Lion Swarm Optimization (LSO) algorithm suffers from two classic pitfalls:

  1. Slow Convergence: It takes too long to find the optimal path.
  2. Local Optima Traps: The algorithm often gets "stuck" in a good-but-not-great solution (a local optimum) and lacks the mechanism to jump out and find the true global best.

In LSO, "cubs" follow the lion king or lionesses to find prey (the optimal solution). In the original version, cub behavior is somewhat "blind"—they choose their movement strategy with equal probability regardless of how the search is progressing. This lack of adaptability leads to inefficient exploration in the early stages and stagnation in the later stages.

Methods

The authors introduce the Adaptive Chaotic Search and Information Entropy LSO (ACIE_LSO), which features three major innovations:

  • Improved Tent Chaotic Mapping: Standard random initialization can lead to uneven population distribution. The authors use a "Tent Map" (a mathematical function that creates chaotic, yet uniformly distributed sequences) and add an adaptive factor to prevent the sequence from getting stuck in "small cycles" or immobile points. This ensures the "lions" start their search covering the entire territory evenly.
  • Information Entropy & Second-Order Norm: To fix the "blind" cub behavior, the authors use Information Entropy to measure the uncertainty and diversity of the population's fitness. This entropy, combined with the second-order norm (a measure of distance), creates a dynamic step-size. Instead of fixed jumps, cubs move based on the current state of the "pride's" knowledge.
  • Adaptive Selection Probability: The algorithm uses a parameter ($\mu$) that changes over time. Early in the simulation, cubs have a higher probability of following adults (exploration). As iterations increase, the probability of cubs being "expelled" to search on their own increases, preventing the group from huddling around a local optimum (exploitation).
  • Adaptive Tent Chaotic Search: At the end of each iteration, the algorithm performs a localized chaotic search around the best individuals to fine-tune the results and help "inert" particles escape local minima.

Datasets

The authors evaluated ACIE_LSO using two distinct approaches:

  1. 18 Benchmark Functions: A suite of standard mathematical "torture tests" for algorithms, including uni-modal (single peak) and multi-modal (many peaks) functions like Sphere, Ackley, and Griewank.
  2. Real-World Regression: The algorithm was used to optimize the weights and thresholds of a Backpropagation (BP) Neural Network to predict house prices using:
  3. Boston House Price Dataset: 506 samples with 14 features.
  4. California House Price Dataset: 20,125 samples with 9 features.

Key Insights and Ideas

  • Chaos as a Tool for Order: While "chaos" sounds negative, in optimization, it provides the ergodicity (the ability to visit every point in a space) that pure randomness lacks. It ensures the algorithm doesn't leave any stone unturned.
  • The Entropy Balance: By using information entropy, the algorithm can "sense" when the population is becoming too similar. If entropy is low, it suggests the lions are converging; the algorithm can then adjust the perturbation to keep the search alive.
  • Dynamic Social Structure: The transition from "following the king" to "being expelled" mimics a natural life cycle that happens to be mathematically optimal for balancing global search (finding the general area of the prey) and local search (pinpointing the exact location).

Key Results

  • Superior Accuracy: ACIE_LSO reached the theoretical optimum ($0$) in 12 out of 18 benchmark functions, significantly outperforming standard LSO, Gray Wolf Optimizer (GWO), and Particle Swarm Optimization (PSO).
  • Stability: The Standard Deviation (SD) of the results was orders of magnitude lower than competitors, meaning the algorithm is highly reliable across different runs.
  • Neural Network Boost: When applied to house price prediction, the ACIE_LSO-optimized neural network achieved the lowest Mean Square Error (MSE). For the Boston dataset, the MSE was 0.0003 compared to 0.0033 for a standard BP network.
  • Faster Convergence: Iteration curves show that ACIE_LSO "drops" to a low error state much earlier than other algorithms, saving computational time.

Limitations and Future Work

Despite the improvements, the authors note a few areas for growth:

  • Complex Multi-Peak Functions: The algorithm still struggles to reach the absolute theoretical optimum on some highly complex, multi-peak functions.
  • Implicit Information: The current model doesn't fully extract the implicit relationships between individuals in the population.
  • Narrow Application: So far, the "optimized neural network" portion has only been tested on standard BP networks. The authors plan to apply ACIE_LSO to more complex architectures like Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks for hyperparameter tuning.

Summary updated

Abstract

Abstract This paper proposes an improved adaptive lion swarm optimization (LSO) algorithm integrating the chaotic search strategy and information entropy to address the problem that the standard LSO algorithm has slow convergence and easily falls into the local optimum in later iterations. At first, an adaptive factor is introduced to improve tent chaotic mapping and used for population position initialization to enhance population diversity and realize uniform traversal while ensuring random distribution, ultimately improving the global search ability. Second, to address the problem that the cub selection strategy is blind, resulting in insufficient traversal in the early stage, a dynamic step-size perturbation factor is established using the second-order norm and information entropy. Adaptive parameters are used to dynamically adjust the selection probability of different cub behaviors based on the number of iterations to suppress the premature convergence of the algorithm. Finally, tent chaotic search is employed to adaptively adjust the search range and improve the individuals with poor fitness through multiple neighborhood points of the local optimal solution, further improving the algorithm’s search speed and accuracy. Experimental results on 18 benchmark functions revealed that the proposed algorithm yields superior performance in terms of convergence speed, optimization accuracy, and ability to jump out of the local optimal solution compared with the standard LSO, gray wolf optimizer, and particle swarm optimization algorithms. Furthermore, the improved LSO algorithm was used to optimize the initial weights and thresholds of the BP neural network, and the effectiveness of the proposed algorithm was further verified by studying the house price prediction problem using two real-world datasets.
BibTeX
@article{liu2023adaptive,
  author = {Miaomiao Liu and Yuying Zhang and Jingfeng Guo and Jing Chen and Zhigang Liu},
  title = {{An Adaptive Lion Swarm Optimization Algorithm Incorporating Tent Chaotic Search and Information Entropy}},
  journal = {International Journal of Computational Intelligence Systems},
  year = {2023},
  doi = {10.1007/s44196-023-00216-7},
}
Figure