(A) fixed-point deep reinforcement learning platform with quantization-aware training and adaptive parallelism동적 비트 양자화와 적응형 병렬화를 이용한 고정 소수점 기반 심층강화학습 가속 플랫폼

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 81
  • Download : 0
DC FieldValueLanguage
dc.contributor.advisorKim, Joo-Young-
dc.contributor.advisor김주영-
dc.contributor.authorYang, Je-
dc.date.accessioned2023-06-26T19:33:47Z-
dc.date.available2023-06-26T19:33:47Z-
dc.date.issued2022-
dc.identifier.urihttp://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=997256&flag=dissertationen_US
dc.identifier.urihttp://hdl.handle.net/10203/309859-
dc.description학위논문(석사) - 한국과학기술원 : 전기및전자공학부, 2022.2,[iv, 25 p. :]-
dc.description.abstractWe present a deep reinforcement learning acceleration platform named FIXAR, which employs fixed-point data types and arithmetic units for the first time using a SW/HW co-design approach. We propose a quantization-aware training algorithm in fixed-point, which enables to reduce the data precision by half after a certain amount of training time without losing accuracy. We also design a FPGA accelerator that employs adaptive dataflow and parallelism to handle both inference and training operations. Its processing element has configurable datapath to efficiently support the proposed quantized-aware training. We validate our FIXAR platform, where the host CPU emulates the DRL environment and the FPGA accelerates the agent’s DNN operations, by running multiple benchmarks in continuous action spaces based on a latest DRL algorithm called DDPG. Finally, the FIXAR platform achieves 25293.3 inferences per second (IPS) training throughput, which is 2.7 times higher than the CPU-GPU platform. In addition, its FPGA accelerator shows 53826.8 IPS and 2638.0 IPS/W energy efficiency, which are 5.5 times higher and 15.4 times more energy efficient than those of GPU, respectively. FIXAR also shows the best IPS throughput and energy efficiency among other state-of-the-art acceleration platforms using FPGA, even it targets one of the most complex DNN models.-
dc.languageeng-
dc.publisher한국과학기술원-
dc.title(A) fixed-point deep reinforcement learning platform with quantization-aware training and adaptive parallelism-
dc.title.alternative동적 비트 양자화와 적응형 병렬화를 이용한 고정 소수점 기반 심층강화학습 가속 플랫폼-
dc.typeThesis(Master)-
dc.identifier.CNRN325007-
dc.description.department한국과학기술원 :전기및전자공학부,-
dc.contributor.alternativeauthor양제-
Appears in Collection
EE-Theses_Master(석사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0