Nano-scale fault tolerant machine learning for cognitive radio

We introduce a machine learning based channel state classifier for cognitive radio, designed for nano-scale implementation. The system uses analog computation, and consists of cyclostationary feature extraction and a radial basis function network for classification. The description of the system is partially abstract, but our design choices are motivated by domain knowledge and we believe the system will be feasible for future nanotechnology implementation. We describe an error model for the system, and simulate experimental performance and fault tolerance of the system in recognizing WLAN signals, under different levels of input noise and computational errors. The system performs well under the expected non-ideal manufacturing and operating conditions.