WEKO3
アイテム
Optical neural network hardware using self-referential holography
http://hdl.handle.net/10228/0002001190
http://hdl.handle.net/10228/000200119082b09856-92fe-4499-a70f-f5fea2aa75d5
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
10443733.pdf (85 KB)
|
|
| アイテムタイプ | 共通アイテムタイプ(1) | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 公開日 | 2025-01-30 | |||||||||||||
| タイトル | ||||||||||||||
| タイトル | Optical neural network hardware using self-referential holography | |||||||||||||
| 言語 | en | |||||||||||||
| 著者 |
高林, 正典
× 高林, 正典
WEKO
35482
× Tomioka, Rio
|
|||||||||||||
| 抄録 | ||||||||||||||
| 内容記述タイプ | Abstract | |||||||||||||
| 内容記述 | Artificial intelligence (AI) hardware has attracted much attention in recent years to address future concerns about implementing AI with Neumann-type computers, namely, increasing computational cost and power consumption. We introduce self-referential holographic deep neural network (SR-HDNN): an optoelectronic deep neural network hardware which enables a highly parallel calculation by optical processing unit and a flexible calculation including a nonlinear processing by electronic processing unit are combined to realize several kinds of neural networks. In the presentation, first, we discuss recent trends on AI hardware based on the spatial modulation of light which have been expected to be a hardware implementation method of high-dimension neural network and/or to reduce dimensions of input data to effectively connect electronically-performed neural network or neural network chip. Next, self-referential holography (SRH) which is the base technology of SR-HDNN and how SR-HDNN works will be explained. Then, the advantages will be discussed. Also, we will show the demonstration results of 4-class MNIST handwritten digit image recognition by SR-HDNN using the numerical simulation method with a transmission matrix which describes relationship of the complex amplitude distributions between arbitrary input and output planes. | |||||||||||||
| 言語 | en | |||||||||||||
| 備考 | ||||||||||||||
| 内容記述タイプ | Other | |||||||||||||
| 内容記述 | The 22nd International Symposium on Eco-materials Processing and Design (ISEPD 2024) ,21-24 January, 2024, Nakhon Ratchasima, Thailand | |||||||||||||
| 言語 | en | |||||||||||||
| 書誌情報 |
p. INV-S8_O5, 発行日 2024-01 |
|||||||||||||
| キーワード | ||||||||||||||
| 主題Scheme | Other | |||||||||||||
| 主題 | Volume holography | |||||||||||||
| キーワード | ||||||||||||||
| 主題Scheme | Other | |||||||||||||
| 主題 | Self-referential holography | |||||||||||||
| キーワード | ||||||||||||||
| 主題Scheme | Other | |||||||||||||
| 主題 | Optoelectronic deep neural network | |||||||||||||
| 言語 | ||||||||||||||
| 言語 | eng | |||||||||||||
| 資源タイプ | ||||||||||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_5794 | |||||||||||||
| 資源タイプ | conference paper | |||||||||||||
| 出版タイプ | ||||||||||||||
| 出版タイプ | VoR | |||||||||||||
| 出版タイプResource | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |||||||||||||
| URI | ||||||||||||||
| 識別子タイプ | URI | |||||||||||||
| 関連識別子 | https://isepd2024.com/ | |||||||||||||
| 会議記述 | ||||||||||||||
| 会議名 | The 22nd International Symposium on Eco-materials Processing and Design (ISEPD 2024) | |||||||||||||
| 言語 | en | |||||||||||||
| 回次 | 22 | |||||||||||||
| 開始年 | 2024 | |||||||||||||
| 開始月 | 01 | |||||||||||||
| 開始日 | 22 | |||||||||||||
| 終了年 | 2024 | |||||||||||||
| 終了月 | 01 | |||||||||||||
| 終了日 | 24 | |||||||||||||
| 開催国 | THA | |||||||||||||
| 査読の有無 | ||||||||||||||
| 値 | yes | |||||||||||||
| 研究者情報 | ||||||||||||||
| URL | https://hyokadb02.jimu.kyutech.ac.jp/html/100000508_ja.html | |||||||||||||
| 論文ID（連携） | ||||||||||||||
| 値 | 10443733 | |||||||||||||
| 連携ID | ||||||||||||||
| 値 | 12566 | |||||||||||||
