We fabricated neural-based superconducting integrated-circuits by using Nb/AlOx/Nb Josephson junctions, and demonstrated circuit operation of 2-bit neural-based A/D converter as an example of optimization problem. We used, in the present superconducting neural circuits, fluxon pulses as the neural impulses and a Josephson junction as a threshold element to achieve an advanced operation. The integrative time delays due to the inductance of superconducting circuits correspond to those due to capacitance in real neurons. The values of resistor by which Josephson transmission lines are connected represent fixed synaptic strengthes. The preliminary experimental result suggests that variable critical currents of d.c. SQUID may provide synapses with variable strength. In our implementation scheme, complex circuit design can easily be realized by connecting Josephson transmission lines. In order to design complex circuits like a Hopfield neural network, we propose a new method which can be used to analyze the properties of local minima in the energy function. A novel design technique to eliminate these local minima in the networks has been developed. It is useful for hardware studies on artificial neural networks using superconductor and semiconductor.
ASJC Scopus subject areas
- Physics and Astronomy(all)