| Presentation | 2020-08-27 [Invited Talk] Recent Advance of Deep-Unfolded Algorithms for Signal Processing and Wireless Communications Satoshi Takabe, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In this talk, I will briefly review recent progress of deep unfolding as a promising deep learning technique. A network architecture of deep unfolding is given by unfolding the recursive structure of existing iterative algorithms and embedding trainable parameters. Then, if all sub-processes of the unfolded algorithm are differentiable, these trainable parameters can be learned by standard deep learning techniques such as back propagation. In this sense, deep unfolding is a variant of data-driven algorithm design and deeply related to recently proposed idea of differentiable programming. Advantages of deep unfolding are as follows: (i) a wide range of domain-specific algorithms are available. (ii) Deep-unfolded algorithms can improve convergence speed and convergent performance by learning rather small number of trainable parameters with low training cost. (iii) Learned parameters are possibly interpretable. From these advantages, deep unfolding has been widely applied to signal processing including wireless communications and other inverse problems. In the talk, first I will introduce a basic idea of deep unfolding, and present some examples of deep-unfolded algorithms for compressed sensing and other wireless communication problems such as MIMO signal detection. We also discuss future directions of deep unfolding. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | deep learning / deep unfolding / MIMO / compressed sensing |
| Paper # | SIP2020-30 |
| Date of Issue | 2020-08-20 (SIP) |
| Conference Information | |
| Committee | SIP |
|---|---|
| Conference Date | 2020/8/27(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Kazunori Hayashi(Kyoto Univ.) |
| Vice Chair | Yukihiro Bandou(NTT) / Toshihisa Tanaka(Tokyo Univ. Agri.&Tech.) |
| Secretary | Yukihiro Bandou(Hosei Univ.) / Toshihisa Tanaka(Waseda Univ.) |
| Assistant | Yuichi Tanaka(Tokyo Univ. Agri.&Tech.) |
| Paper Information | |
| Registration To | Technical Committee on Signal Processing |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | [Invited Talk] Recent Advance of Deep-Unfolded Algorithms for Signal Processing and Wireless Communications |
| Sub Title (in English) | |
| Keyword(1) | deep learning |
| Keyword(2) | deep unfolding |
| Keyword(3) | MIMO |
| Keyword(4) | compressed sensing |
| 1st Author's Name | Satoshi Takabe |
| 1st Author's Affiliation | Nagoya Institute of Technology(NITech) |
| Date | 2020-08-27 |
| Paper # | SIP2020-30 |
| Volume (vol) | vol.120 |
| Number (no) | SIP-142 |
| Page | pp.pp.11-11(SIP), |
| #Pages | 1 |
| Date of Issue | 2020-08-20 (SIP) |