Topic & Info
CAD & SoC Design Laboratory
Deep Learning Hardware

It is estimated that the human brain consists of approximately 25 billion nerve cells. Artificial neural network (ANN) is a mathematical model that imitates human brain. The human brain can categorize the data it has received. It was created through a long experience. Similarly, ANN can create algorithms that can be classified through data, and the process of creating it is called "learning". The single layer neural network, the simplest model of ANN, is shown in Fig. In this figure, w is the value learned through learning. Data can be categorized through output made using input data and weight. That increase the number of nodes with a layer to improve the performance of such a change the ANN algorithm is called deep learning.

Currently, Deep learning is implemented in software. However, it is not suitable for mobile devices because it requires CPU and GPU operation. A neuromorphic chip can solve this problem. The neuromorphic chip is a chip that can perform such deep learning. This chip has better performance and lower power consumption compared to conventional methods. Therefore, by using neuromorphic chip, artificial intelligence using mobile device will be further developed.