Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learning

Meta-learning has been the most common framework for few-shot learning in recent years. It learns the model from collections of few-shot classification tasks, which is believed to have a key advantage of making the training objective consistent with the testing objective. However, some recent works...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Proceedings / IEEE International Conference on Computer Vision s. 9042 - 9051
Hlavní autoři: Chen, Yinbo, Liu, Zhuang, Xu, Huijuan, Darrell, Trevor, Wang, Xiaolong
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 01.10.2021
Témata:
Classification algorithms
Codes
Computer vision
Measurement
Task analysis
Training
Transfer/Low-shot/Semi/Unsupervised Learning
ISSN:2380-7504
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Meta-learning has been the most common framework for few-shot learning in recent years. It learns the model from collections of few-shot classification tasks, which is believed to have a key advantage of making the training objective consistent with the testing objective. However, some recent works report that by training for whole-classification, i.e. classification on the whole label-set, it can get comparable or even better embedding than many meta-learning algorithms. The edge between these two lines of works has yet been underexplored, and the effectiveness of meta-learning in few-shot learning remains unclear. In this paper, we explore a simple process: meta-learning over a whole-classification pre-trained model on its evaluation metric. We observe this simple method achieves competitive performance to state-of-the-art methods on standard bench-marks. Our further analysis shed some light on understanding the trade-offs between the meta-learning objective and the whole-classification objective in few-shot learning. Our code is available at https://github.com/yinboc/few-shot-meta-baseline.
ISSN:2380-7504
DOI:10.1109/ICCV48922.2021.00893