Performance of Ag–Ag2S core–shell nanoparticle-based random network reservoir computing device

• Hadiyawarman

• 宇佐美, 雄生

• Kotooka, Takumi

• Azhari, Saman

• Eguchi, Masanori

• 田中, 啓文

抄録

Reservoir computing (RC), a low-power computational framework derived from recurrent neural networks, is suitable for temporal/sequential data processing. Here, we report the development of RC devices utilizing Ag–Ag2S core–shell nanoparticles (NPs), synthesized by a simple wet chemical protocol, as the reservoir layer. We examined the NP-based reservoir layer for the required properties of RC hardware, such as echo state property, and then performed the benchmark tasks. Our study on NP-based reservoirs highlighted the importance of the dynamics between the NPs as indicated by the rich high dimensionality due to the echo state property. These dynamics affected the accuracy (up to 99%) of the target waveforms that were generated with a low number of readout channels. Our study demonstrates the great potential of Ag–Ag2S NPs for the development of next-generation RC hardware.

収録刊行物

• Japanese Journal of Applied Physics

  Japanese Journal of Applied Physics 60 (SC), SCCF02-1-SCCF02-5, 2021-02-22

  応用物理学会

キーワード

• General Physics and Astronomy
• Physics and Astronomy (miscellaneous)
• General Engineering

詳細情報

• CRID: 1050010293092802688
• NII論文ID: 120007193028
• NII書誌ID: AA12295836
• ISSN: 13474065
• ISSN: 00214922
• DOI: 10.35848/1347-4065/abe206
• HANDLE: 10228/00008735
• Web Site: https://kyutech.repo.nii.ac.jp/records/7531
• Web Site: https://iopscience.iop.org/article/10.35848/1347-4065/abe206
• Web Site: https://iopscience.iop.org/article/10.35848/1347-4065/abe206/pdf
• 本文言語コード: en
• 資料種別: journal article
• データソース種別:
  • IRDB
  • Crossref
  • CiNii Articles
  • KAKEN