Résumé
Compact oscillating devices can be realized by exploiting the insulator-to-metal transition of VO2 (IMT). Coupled oscillators can be locked in frequency and phase through connections realized with resistors or capacitors. Oscillatory neural networks based on VO2 devices are researched as neuromorphic computing hardware for image recognition applications. Physical demonstrations of large networks of coupled oscillators are yet hindered by the device-to-device variability. In this work, we present a frequency-injection scheme which allows additional tuning and control of the oscillator network comprising non-ideal VO2 devices.