Nota, Inc.

Jun Kyeong Choi

Shin Kook Choi

Provided are a method and apparatus for performing a convolution operation for optimizing the arithmetic intensity of device. The method includes obtaining input data, and generating output data through a convolution model for a device, by using the input data, wherein the convolution model includes a plurality of convolutional layers, the plurality of convolutional layers include an improved first convolutional layer that satisfies preset conditions related to latency characteristics of the device, the improved first convolutional layer is a layer obtained by adjusting a first convolutional layer so as to perform a first convolution operation on a first feature map, which is obtained by reshaping an input feature map of the first convolutional layer based on a predetermined final division value, and the final division value is determined considering the latency characteristics such that an overall arithmetic intensity of the improved first convolutional layer is increased compared to the first convolutional layer.

Growing community of inventors

Seoul, South Korea

Jun Kyeong Choi