XCS-CR for handling input, output, and reward noise
Takato Tatsumi, Keiki Takadama · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2019
To briefly represent a dataset, it is crucial to find common attributes among the data. Extended learning classifier system (XCS) finds common attributes of multiple data and acquires generalized rules that match multiple data. In real-world problems, it may be challenging to find common attributes due to noise in the data and the inability of XCS to acquire the generalized rules. Considering a classification problem as an example, noise may be included at each input, output, as well as in the evaluation of the output. To tackle this problem, our previous work proposed XCSs that acquire appropriately generalized rules, specifically for a problem with one of the three mentioned type of noises added. In real-world problems, it is difficult to identify the type of noise in advance, which requires an XCS to cope with multiple types of noise. For this issue, this paper proposes an XCS that can handle any noise on the input, output, and evaluation of the output, and aims at investigating the effectiveness of the proposed XCS in the multiplexer problems including any of the three types of noise.