The article discusses a new artificial synapse technology based on superconducting Josephson junctions with magnetic clusters. This technology aims to create more energy-efficient neuromorphic computing systems.
Key points:
– The artificial synapse consists of axonal and dendritical superconducting electrodes with a magnetic barrier in between
– The magnetic barrier contains tunable magnetic clusters that can switch between ordered and disordered states
– The critical current of the junction, analogous to synaptic weight, can be adjusted using input voltage spikes
– Numerical modeling shows the devices can operate in a stochastic regime with spiking energy comparable to thermal energy
– The technology is more energy-efficient than current neuromorphic hardware, even when considering cooling requirements
– This could enable more complex computing systems due to improved energy efficiency
The article proposes a novel artificial synapse design using superconducting Josephson junctions with magnetic clusters. This technology aims to create more energy-efficient neuromorphic computing systems by allowing the critical current of the junction to be tuned using input voltage spikes. Numerical modeling indicates the devices can operate in a stochastic regime with spiking energy comparable to thermal energy, making them more energy-efficient than current neuromorphic hardware. This could ultimately enable more complex computing systems due to improved energy efficiency.
Source: https://www.nist.gov/patents/neural-member-neural-network-and-neurological-memory
Keywords: 1. Synapse, 2. Superconducting, 3. Josephson, 4. Magnetic, 5. Neuromorphic
