The interaction of a gas particle with a metal-oxide based gas sensor changes the sensor irreversibly. The compounded changes, referred to as sensor drift, are unstable, but adaptive algorithms can sustain the accuracy of odor sensor systems. This paper shows how such a system can be defined without additional data acquisition by transfering knowledge from one time window to a subsequent one after drift has occurred. A context-based neural network model is used to form a latent representation of sensor state, thus making it possible to generalize across a sequence of states. When tested on samples from unseen subsequent time windows, the approach performed better than drift-naive and ensemble methods on a gas sensor array drift dataset. By reducing the effect that sensor drift has on classification accuracy, context-based models may be used to extend the effective lifetime of gas identification systems in practical settings.

PDF

arXiv:2003.07292, 2020.

@article{warner:arxiv20, title={Using context to make gas classifiers robust to sensor drift}, author={Jamieson Warner and Ashwin Devaraj and Risto Miikkulainen}, journal={arXiv:2003.07292}, month={ }, url="http://nn.cs.utexas.edu/?warner:arxiv20", year={2020} }