Design and implementation of an ultra high speed and low energy artificial neuron

The current scientific community captivated by understanding the general principles of human brain functions, as a further matter, on how to mimic the abilities by utilizing artificial neurons for more efficient computing. Functionality comes from self-assembly of brain cells, known as nerve cells or neurons. Modeling human brain cell at a molecular level is not practical on account of its biological complexity. Mathematical approaches and technological developments led the hardware and software implementation of artificial neuron models easier. Computational software tools connect artificial neurons to each other to create Artificial Neural Network (ANN) to adopt biological neural network behavior. ANN software tools have gained extensive acceptance for wide range use of neural network applications because of learning abilities, computational power and speed through parallel processing. Furthermore, the models of ANN are simpler than conventional computing devices. For high performance computing circuits, a neuron cell can enhance the possibilities of solving complex problems. Therefore, a basic neuron model has the capacity of building an ANN on chip or hybrid digital circuits. vii To mimic the biological brain cell, this study shows a Josephson Junction (JJ) based Artificial Neuron (JJ-AN) circuit that satisfies the capability of creating ANN on chip and leaky Integrate and Fire Neuron (IFN) model. This artificial neuron circuit is formed by three main structures: a double-junction SQUID interfered with a resistor (threshold loop), adjoined resistor and inductance structure (decaying loop), and mutual conductance between threshold loop and decaying loop inductances. The proposed model has only one input and one output ports and it makes the circuit relatively simple. Nevertheless, optimization of such a design is a crucial process as neuron circuit is not only used together with other neuron circuits but also combined all together with Single Flux Quantum (SFQ) digital library circuits. Operation frequency of neuron model is observed up to 120 GHz. As an example for the research, two parameter sets that make different threshold values are converged by the modified optimizer that shows minimum margins of ±23% and ±7% respectively.