Apparatus for solving circuit equations of processing elements using neural network and method for controlling the same

A method for solving circuit equations of a processing element (PE) using a neural network, the method being performed by a graphic processing unit (GPU), the method comprising:

• forming, in the neural network, M aligned virtual cell arrays, each of the M aligned virtual cell arrays comprising a plurality of memory cells, wherein the M aligned virtual cell arrays have a height value N and a width value O, and each of the M aligned virtual cell arrays corresponds to a crossbar array circuit included in the processing element;
• performing a 3D convolution on each of the M aligned virtual cell arrays until the height value N of each of the M aligned virtual cell arrays becomes 1;
• inputting parameters of the plurality of memory cells of the crossbar array circuit to each of the M aligned virtual cell arrays in which the 3D convolution is performed;
• and
• solving a circuit equation of the processing element using an output value generated from the M aligned virtual cell arrays.