http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Priority Encoder를 활용한 DTCAM-based PIM architecture 설계
이언경(Eon-Gyeong Lee),김경민(Kyeong-Min Kim),박성배(Seong-Bae Park),홍충선(Choong-Seon Hong),홍상훈(Sang-Hoon Hong) 대한전자공학회 2022 대한전자공학회 학술대회 Vol.2022 No.11
This paper describes the DTCAM structure that has increased utilization in artificial intelligence search. The inverter chain of the existing DTCAM structure was improved to reduce power consumption, and a detection module was added to facilitate data retrieval. In addition, it aims to accelerate artificial intelligence computation using memory through a Priority Encoder structure that sets priority to multiple searched memory addresses and outputs them. This is a method of solving the von Neumann bottleneck through PIM, and has various possibilities of expansion in the field of future artificial intelligence computation.
패턴 학습이 가능한 TCAM 메모리를 이용한 비지도 학습
박소희(So-Hee Park),방민경(Min-kyung Bang),이언경(Eon-gyeong Lee),한재훈(Jae-Hun Han),구원모(Won-mo Koo),박성배(Seong-Bae Park),홍충선(Choong-Seon Hong),홍상훈(Sang-Hoon Hong) 대한전자공학회 2023 대한전자공학회 학술대회 Vol.2023 No.6
Processing in memory (PIM) is an attention drawing technology that provides a solution to the von Neumann bottleneck, which occurs in data transmissions between the CPU and memory, providing massive improvement in the overall data processing performance. In this paper, we propose unsupervised learning algorithm using the PIM artificial intelligence processors over TCAM. By using the data preprocessing of thinning and thickening, the MNIST data has clustered with 88.74% accuracy. It displays an advantage in that it consists of high accuracy even with a simple algorithm.