| Presentation | 2022-12-16 Sampling Strategy in Data Pruning Ryota Higashi, Toshikazu Wada, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Data Pruning is a method of selecting the training data out of an entire training dataset so as to keep the accuracy after training. In discriminative models, Hard Examples (HEs) that are close to the decision boundary is important for training. However, when HEs are selected based on pretrained model’s outputs, the accuracy gets worse than random sampling as the sample size decreases. This phenomenon is caused that feature space cannot be reconstructed only from HE, but it is unclear if the phenomenon occurs by other criteria. The experiments using selected datasets by different criteria showed that the appropriate criterion varies by sample size, and it is necessary to combine different criteria in Data Pruning. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Deep Learning / Pruning / Image Classification / Decision Boundary / Sampling Strategy |
| Paper # | PRMU2022-48 |
| Date of Issue | 2022-12-08 (PRMU) |
| Conference Information | |
| Committee | PRMU |
|---|---|
| Conference Date | 2022/12/15(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Toyama International Conference Center |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Seiichi Uchida(Kyushu Univ.) |
| Vice Chair | Takuya Funatomi(NAIST) / Mitsuru Anpai(Denso IT Lab.) |
| Secretary | Takuya Funatomi(CyberAgent) / Mitsuru Anpai(Univ. of Tokyo) |
| Assistant | Nakamasa Inoue(Tokyo Inst. of Tech.) / Yasutomo Kawanishi(Riken) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Sampling Strategy in Data Pruning |
| Sub Title (in English) | |
| Keyword(1) | Deep Learning |
| Keyword(2) | Pruning |
| Keyword(3) | Image Classification |
| Keyword(4) | Decision Boundary |
| Keyword(5) | Sampling Strategy |
| 1st Author's Name | Ryota Higashi |
| 1st Author's Affiliation | Wakayama University(Wakayama Univ.) |
| 2nd Author's Name | Toshikazu Wada |
| 2nd Author's Affiliation | Wakayama University(Wakayama Univ.) |
| Date | 2022-12-16 |
| Paper # | PRMU2022-48 |
| Volume (vol) | vol.122 |
| Number (no) | PRMU-314 |
| Page | pp.pp.85-90(PRMU), |
| #Pages | 6 |
| Date of Issue | 2022-12-08 (PRMU) |