In his talk, Victor presented the promising potential of memristive devices in neuromorphic systems. He began by explaining the key features of memristor architecture, materials, production technologies, and properties. Victor then compared neuromorphic systems with traditional electronic circuits, highlighting challenges such as noise and cross-talk, which are detrimental in conventional circuits but play a positive role in neuromorphic systems. He discussed the difficulties in hardware realization of “classic” artificial neural networks (perceptrons) and the limitations of software-based implementations, particularly regarding power efficiency. Victor also explored several artificial neural networks realized on memristive devices, such as perceptrons, spiking neuron networks, and reservoir computing, evaluating their advantages and drawbacks. He introduced a memristive circuit mimicking Pavlov’s dog learning through the STDP (Spike Timing Dependent Plasticity) algorithm and discussed the potential for neuro-prostheses, particularly for patients with spinal cord injuries.
This talk by Victor Erokhin was held on November 20th at DSC EUROPE 24 in Belgrade.
Follow us on social media :
LinkedIn: https://www.linkedin.com/company/11184830/admin/
Instagram: https://www.instagram.com/datasciconf/
Facebook page: https://www.facebook.com/DataSciConference
Website: https://datasciconference.com/
