VALIS, a Novel Immune-inspired Supervised Learning Algorithm with Applications to Soft Measurements

We present VALIS (Vote-Allocating Immune System), a supervised learning algorithm based on the principles of the immune system. Classification is performed by a population of artificial antibodies that can bind to the input data and vote for their classes. The training is performed in evolutionary manner, with new antibodies being created by means of crossover and mutation. Tests on multiple problems demonstrate that VALIS is competitive with established classification methods. Based on VALIS, a system protecting the wireless sensor networks from attacks by mutating viruses can be developed. To this end, intelligent (soft) sensors which can be trained by the evolutionary algorithm to determine the type of attack and neutralize it are employed. The machine learning technology considered belongs to the area of soft measurement methods.