Cancer model identification via sliding mode and differential neural networks

The present paper provides a description for the identification process of the cancer mathematical model proposed by Lopez and Marco under the immunotherapy treatment by differential neural networks and sliding mode type observer techniques. The combination of these both techniques make available a close enough tracking between the estimate states given by the neural network and the cancer model dynamics: these are the interleukin-2, the tumor cells and the effector cells concentrations. The feedback error and the sign function error are the hints for application into the learning algorithm. This algorithm is tested by numerical calculations and at the same time, it looks as an important opportunity to build feedbacks controls.