Loss Architecture Search for Few-Shot Object Recognition
Few-shot object recognition, which exploits a set of well-labeled data to build a classifier for new classes that have only several samples per class, has received extensive attention from the machine learning community. In this paper, we investigate the problem of designing an optimal loss function...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2020-01-01
|
| Series: | Complexity |
| Online Access: | http://dx.doi.org/10.1155/2020/1041962 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832566335302795264 |
|---|---|
| author | Jun Yue Zelang Miao Yueguang He Nianchun Du |
| author_facet | Jun Yue Zelang Miao Yueguang He Nianchun Du |
| author_sort | Jun Yue |
| collection | DOAJ |
| description | Few-shot object recognition, which exploits a set of well-labeled data to build a classifier for new classes that have only several samples per class, has received extensive attention from the machine learning community. In this paper, we investigate the problem of designing an optimal loss function for few-shot object recognition and propose a novel few-shot object recognition system that includes the following three steps: (1) generate a loss function architecture using a recurrent neural network (generator); (2) train a base embedding network with the generated loss function on a training set; (3) fine-tune the base embedding network using the few-shot instances from a validation set to obtain the accuracy and use it as a reward signal to update the generator. This procedure is repeated and implemented in the reinforcement learning framework for finding the best loss architecture such that the embedding network yields the highest validation accuracy. Our key insight is to create a search space of the loss function architectures and evaluate the quality of a particular loss function on the dataset of interest. We conduct experiments on three popular datasets for few-shot learning. The results show that the proposed approach achieves better performance than state-of-the-art methods. |
| format | Article |
| id | doaj-art-5186599e3ef640048d654eccfb586fb1 |
| institution | Kabale University |
| issn | 1076-2787 1099-0526 |
| language | English |
| publishDate | 2020-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | Complexity |
| spelling | doaj-art-5186599e3ef640048d654eccfb586fb12025-02-03T01:04:28ZengWileyComplexity1076-27871099-05262020-01-01202010.1155/2020/10419621041962Loss Architecture Search for Few-Shot Object RecognitionJun Yue0Zelang Miao1Yueguang He2Nianchun Du3Department of Geomatics Engineering, School of Traffic & Transportation Engineering, Changsha University of Science & Technology, Changsha, ChinaSchool of Geosciences and Info-Physics, Central South University, Changsha, ChinaDepartment of Geomatics Engineering, School of Traffic & Transportation Engineering, Changsha University of Science & Technology, Changsha, ChinaChina Nonferrous Metal Changsha Survey and Design Institute, Changsha, ChinaFew-shot object recognition, which exploits a set of well-labeled data to build a classifier for new classes that have only several samples per class, has received extensive attention from the machine learning community. In this paper, we investigate the problem of designing an optimal loss function for few-shot object recognition and propose a novel few-shot object recognition system that includes the following three steps: (1) generate a loss function architecture using a recurrent neural network (generator); (2) train a base embedding network with the generated loss function on a training set; (3) fine-tune the base embedding network using the few-shot instances from a validation set to obtain the accuracy and use it as a reward signal to update the generator. This procedure is repeated and implemented in the reinforcement learning framework for finding the best loss architecture such that the embedding network yields the highest validation accuracy. Our key insight is to create a search space of the loss function architectures and evaluate the quality of a particular loss function on the dataset of interest. We conduct experiments on three popular datasets for few-shot learning. The results show that the proposed approach achieves better performance than state-of-the-art methods.http://dx.doi.org/10.1155/2020/1041962 |
| spellingShingle | Jun Yue Zelang Miao Yueguang He Nianchun Du Loss Architecture Search for Few-Shot Object Recognition Complexity |
| title | Loss Architecture Search for Few-Shot Object Recognition |
| title_full | Loss Architecture Search for Few-Shot Object Recognition |
| title_fullStr | Loss Architecture Search for Few-Shot Object Recognition |
| title_full_unstemmed | Loss Architecture Search for Few-Shot Object Recognition |
| title_short | Loss Architecture Search for Few-Shot Object Recognition |
| title_sort | loss architecture search for few shot object recognition |
| url | http://dx.doi.org/10.1155/2020/1041962 |
| work_keys_str_mv | AT junyue lossarchitecturesearchforfewshotobjectrecognition AT zelangmiao lossarchitecturesearchforfewshotobjectrecognition AT yueguanghe lossarchitecturesearchforfewshotobjectrecognition AT nianchundu lossarchitecturesearchforfewshotobjectrecognition |