Machine Learning Methods
Sparrow search algorithm outperforms three existing optimization algorithms in benchmark tests
In a 2020 paper, Jiankai Xue and Bo Shen at Donghua University present a new method called the sparrow search algorithm. It is a swarm-intelligence method, meaning it copies the way groups of animals search for food or avoid danger. The authors tested it on 19 benchmark functions, standard test problems that let researchers compare search methods under the same conditions. They compared it with grey wolf optimizer, gravitational search algorithm, and particle swarm optimization. The simulation results rank the sparrow algorithm ahead in accuracy, convergence speed, stability, and robustness. The paper appeared in Systems Science & Control Engineering.
How does the sparrow search algorithm work?
The authors modeled the method on group wisdom, foraging, and anti-predation behavior in sparrows. Those behaviors guide the algorithm as it searches possible answers.
How did the researchers test it?
The authors ran the algorithm on 19 benchmark functions. They compared its performance with three other methods: grey wolf optimizer, gravitational search algorithm, and particle swarm optimization. The tests measured accuracy, convergence speed, stability, and robustness.
What did the results show?
The simulation results show the sparrow search algorithm was superior to grey wolf optimizer, gravitational search algorithm, and particle swarm optimization in accuracy, convergence speed, stability, and robustness. The authors also report effectiveness in two practical engineering examples. The abstract does not give details about those examples.
What is optimization?
Optimization is the task of finding the best answer among many possible choices, such as the lowest cost design or the fastest route. Nature-inspired algorithms give researchers one way to search large sets of possible answers.
What are benchmark functions?
Benchmark functions are standard mathematical test problems with known results. Researchers use them to measure how a new search method performs before trying it on real-world problems.
Why compare with grey wolf optimizer, gravitational search algorithm, and particle swarm optimization?
These are existing optimization methods. Comparing the new method with them gives a reference for how much improvement it offers on the same test problems.
Limits and what's next
The abstract describes a simulation study on 19 benchmark functions, with two practical engineering examples. The setting is limited to those benchmark problems and the two examples. The abstract reports no independent replication by another research team. It also does not say whether the two practical examples were run as simulations or under field conditions. This means the finding rests on this one study, and the abstract does not say how the method would perform on optimization problems outside the tests.
Source: Jiankai Xue and Bo Shen, “A novel swarm intelligence optimization approach: sparrow search algorithm,” Systems Science & Control Engineering, 2020.