WEKO3
アイテム
Field Programmable Gate Arrayでの実装に適した混合精度重みモデルに基づくニューラルネットワーク
https://doi.org/10.18997/00008662
https://doi.org/10.18997/00008662ab5431a9-33e8-4fde-b2c6-10a77259634d
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
|
|
|
| アイテムタイプ | 学位論文 = Thesis or Dissertation(1) | |||||||
|---|---|---|---|---|---|---|---|---|
| 公開日 | 2021-12-20 | |||||||
| 資源タイプ | ||||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_db06 | |||||||
| 資源タイプ | doctoral thesis | |||||||
| タイトル | ||||||||
| タイトル | Mixed-precision weights network for field-programmable gate array | |||||||
| 言語 | en | |||||||
| タイトル | ||||||||
| タイトル | Field Programmable Gate Arrayでの実装に適した混合精度重みモデルに基づくニューラルネットワーク | |||||||
| 言語 | ja | |||||||
| 言語 | ||||||||
| 言語 | eng | |||||||
| 著者 |
Ninnart, Fuengfusin
× Ninnart, Fuengfusin
|
|||||||
| 抄録 | ||||||||
| 内容記述タイプ | Abstract | |||||||
| 内容記述 | In this study, I introduced a mixed-precision weights network (MPWN), which is a quantization neural network that jointly utilizes three different weight spaces: binary {-1, 1}, ternary {-1, 0, 1}, and 32-bit floating-point. I further developed the MPWN from both software and hardware aspects. From the software aspect, I evaluated the MPWN on the Fashion-MNIST, CIFAR10, and ILSVRC 2012 datasets. I systematized the accuracy sparsity bit score, which is a linear combination of accuracy, sparsity, and number of bits. This score allows Bayesian optimization to be used efficiently to search for MPWN weight space combinations. From the hardware aspect, I proposed XOR signed-bits to explore floating-point and binary weight spaces in the MPWN. XOR signed-bits is an efficient implementation equivalent to the multiplication of floating-point and binary weight spaces. Using the concept from XOR signed bits, I also provide a ternary bitwise operation that is an efficient implementation equivalent to the multiplication of floating-point and ternary weight space. To demonstrate the compatibility of the MPWN with hardware implementation, I synthesized and implemented the MPWN in a field-programmable gate array using high-level synthesis. My proposed MPWN implementation utilized up to 1.68-4.89 times less hardware resources depending on the type of resources than a conventional the 32-bit floating-point model. In addition, my implementation reduced the latency up to 31.55 times compared to 32-bit floating-point model without optimizations. | |||||||
| 目次 | ||||||||
| 内容記述タイプ | TableOfContents | |||||||
| 内容記述 | 1 Introduction||2 Background||3 Mixed Precision Weight Network and FPGA Design||4 Related Works||5 Experimental Results and Discussion||6 Conclusion | |||||||
| 備考 | ||||||||
| 内容記述タイプ | Other | |||||||
| 内容記述 | 九州工業大学博士学位論文 学位記番号:生工博甲第419号 学位授与年月日:令和3年9月24日 | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Deep Learning | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | FPGA | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Quantization Neural Networks | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Neural network | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Image recognition | |||||||
| アドバイザー | ||||||||
| 田向, 権 | ||||||||
| 学位授与番号 | ||||||||
| 学位授与番号 | 甲第419号 | |||||||
| 学位名 | ||||||||
| 学位名 | 博士(工学) | |||||||
| 学位授与年月日 | ||||||||
| 学位授与年月日 | 2021-09-24 | |||||||
| 学位授与機関 | ||||||||
| 学位授与機関識別子Scheme | kakenhi | |||||||
| 学位授与機関識別子 | 17104 | |||||||
| 学位授与機関名 | 九州工業大学 | |||||||
| 学位授与年度 | ||||||||
| 内容記述タイプ | Other | |||||||
| 内容記述 | 令和3年度 | |||||||
| 出版タイプ | ||||||||
| 出版タイプ | VoR | |||||||
| 出版タイプResource | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |||||||
| アクセス権 | ||||||||
| アクセス権 | open access | |||||||
| アクセス権URI | http://purl.org/coar/access_right/c_abf2 | |||||||
| ID登録 | ||||||||
| ID登録 | 10.18997/00008662 | |||||||
| ID登録タイプ | JaLC | |||||||