Binary Invasive Weed Optimization

Recently, a new evolutionary algorithm for optimization in continuous spaces called Invasive Weed Optimization (IWO) was introduced. Since IWO employs a real-valued vector representation, the question arises whether it can also be used for problem domains that need a binary encoding. This paper introduces a binary IWO (BinIWO) concept in which the weeds and seeds are defined as bitstrings. The reproduction operation determines the offspring in a normally distributed neighborhood in the space of bitstrings. Thereby, the normal distribution is not defined over the bitstrings, but over the number of bits to be different in the offspring. BinIWO is applied to four typical benchmark functions known from literature and exhibits promising results.