Exploiting Evolution for an Adaptive Drift-Robust Classifier in Chemical Sensing

Gas chemical sensors are strongly affected by drift, i.e., changes in sensors’ response with time, that may turn statistical models commonly used for classification completely useless after a period of time. This paper presents a new classifier that embeds an adaptive stage able to reduce drift effects. The proposed system exploits a state-of-the-art evolutionary strategy to iteratively tweak the coefficients of a linear transformation able to transparently transform raw measures in order to mitigate the negative effects of the drift. The system operates continuously. The optimal correction strategy is learnt without a-priori models or other hypothesis on the behavior of physical-chemical sensors. Experimental results demonstrate the efficacy of the approach on a real problem.