S.A. Gerasimova

An optimal LSTM neural network architecture has been developed and implemented, capable of effectively predicting seizure activity. The obtained architecture has been trained on both human EEG data and rodent EEG data. Memristive neuron-like generators have been developed and studied, demonstrating biologically plausible properties. A neuron-like memristive system for adaptive stimulation of rat brain neurons has been designed and the closed-loop system.