On the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem.
As a promising research direction in recent decades, active learning allows an oracle to assign labels to typical examples for performance improvement in learning systems. Existing works mainly focus on designing criteria for screening examples of high value to be labeled in a handcrafted manner. In...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Public Library of Science (PLoS)
2019-01-01
|
| Series: | PLoS ONE |
| Online Access: | https://doi.org/10.1371/journal.pone.0217408 |
| id |
doaj-38c11e8c6f284674b3e4c329e9955dee |
|---|---|
| record_format |
Article |
| spelling |
doaj-38c11e8c6f284674b3e4c329e9955dee2021-03-03T20:37:22ZengPublic Library of Science (PLoS)PLoS ONE1932-62032019-01-01146e021740810.1371/journal.pone.0217408On the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem.Honglan HuangJincai HuangYanghe FengJiarui ZhangZhong LiuQi WangLi ChenAs a promising research direction in recent decades, active learning allows an oracle to assign labels to typical examples for performance improvement in learning systems. Existing works mainly focus on designing criteria for screening examples of high value to be labeled in a handcrafted manner. Instead of manually developing strategies of querying the user to access labels for the desired examples, we utilized the reinforcement learning algorithm parameterized with the neural network to automatically explore query strategies in active learning when addressing stream-based one-shot classification problems. With the involvement of cross-entropy in the loss function of Q-learning, an efficient policy to decide when and where to predict or query an instance is learned through the developed framework. Compared with a former influential work, the advantages of our method are demonstrated experimentally with two image classification tasks, and it exhibited better performance, quick convergence, relatively good stability and fewer requests for labels.https://doi.org/10.1371/journal.pone.0217408 |
| collection |
DOAJ |
| language |
English |
| format |
Article |
| sources |
DOAJ |
| author |
Honglan Huang Jincai Huang Yanghe Feng Jiarui Zhang Zhong Liu Qi Wang Li Chen |
| spellingShingle |
Honglan Huang Jincai Huang Yanghe Feng Jiarui Zhang Zhong Liu Qi Wang Li Chen On the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem. PLoS ONE |
| author_facet |
Honglan Huang Jincai Huang Yanghe Feng Jiarui Zhang Zhong Liu Qi Wang Li Chen |
| author_sort |
Honglan Huang |
| title |
On the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem. |
| title_short |
On the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem. |
| title_full |
On the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem. |
| title_fullStr |
On the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem. |
| title_full_unstemmed |
On the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem. |
| title_sort |
on the improvement of reinforcement active learning with the involvement of cross entropy to address one-shot learning problem. |
| publisher |
Public Library of Science (PLoS) |
| series |
PLoS ONE |
| issn |
1932-6203 |
| publishDate |
2019-01-01 |
| description |
As a promising research direction in recent decades, active learning allows an oracle to assign labels to typical examples for performance improvement in learning systems. Existing works mainly focus on designing criteria for screening examples of high value to be labeled in a handcrafted manner. Instead of manually developing strategies of querying the user to access labels for the desired examples, we utilized the reinforcement learning algorithm parameterized with the neural network to automatically explore query strategies in active learning when addressing stream-based one-shot classification problems. With the involvement of cross-entropy in the loss function of Q-learning, an efficient policy to decide when and where to predict or query an instance is learned through the developed framework. Compared with a former influential work, the advantages of our method are demonstrated experimentally with two image classification tasks, and it exhibited better performance, quick convergence, relatively good stability and fewer requests for labels. |
| url |
https://doi.org/10.1371/journal.pone.0217408 |
| work_keys_str_mv |
AT honglanhuang ontheimprovementofreinforcementactivelearningwiththeinvolvementofcrossentropytoaddressoneshotlearningproblem AT jincaihuang ontheimprovementofreinforcementactivelearningwiththeinvolvementofcrossentropytoaddressoneshotlearningproblem AT yanghefeng ontheimprovementofreinforcementactivelearningwiththeinvolvementofcrossentropytoaddressoneshotlearningproblem AT jiaruizhang ontheimprovementofreinforcementactivelearningwiththeinvolvementofcrossentropytoaddressoneshotlearningproblem AT zhongliu ontheimprovementofreinforcementactivelearningwiththeinvolvementofcrossentropytoaddressoneshotlearningproblem AT qiwang ontheimprovementofreinforcementactivelearningwiththeinvolvementofcrossentropytoaddressoneshotlearningproblem AT lichen ontheimprovementofreinforcementactivelearningwiththeinvolvementofcrossentropytoaddressoneshotlearningproblem |
| _version_ |
1714821496989286400 |