ACC
Graph-based Embedding Propagation for Transductive Few-shot Classification
TOTAL VIEWS: 1021
Abstract
The purpose of few-shot learning is to identify underlying patterns in data using only a limited number of training samples. Few-shot classification aims to train a classifier to differentiate between previously unseen classes based on a small number of labeled instances. Few-shot classification is highly difficult due to unseen classes and data-poor issues. Extraction of features that are typical of each category is, in our opinion, important for classification tasks. However, many existing approaches extract features directly from a small number of samples inside each class without taking into account the relationship information between them, leaving the classes with insufficient discriminative features. In this paper, I propose novel extensions of the Transductive Propagation Network (TPN), called Co-Attention TPN (CA-TPN) which is a method to extract representative feature information by considering the relational information among input data channels and class samples. Additionally, the study conducts semi-supervised few-shot classification through graph-based embedding propagation. First, this study focuses on examining the correlation between input data channels and developing a metric score formula to assess the significance of each channel feature. This approach involves incorporating a channel attention mechanism that enables us to weigh the relevance of each channel feature accurately. Second, I explore the relationship between the samples in each class and propose a novel prototype model that utilizes relation attention. This prototype with relation attention serves as a replacement for the traditional embedding feature of a single sample and enhances subsequent embedding propagation and downstream classification tasks. The approach I employ is validated on various benchmark datasets, demonstrating superior performance compared to existing few-shot classification methods and attaining the current state-of-the-art outcomes.
References
[1] Wang, Y., Yao, Q., Kwok, J. T., & Ni, L. M. (2020). Generalizing from a few examples: A survey on few-shot learning. Acm Computing Surveys (CSUR), 53(3), 1-34.
[2] Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017). Imagenet classification with deep convolutional neural networks. Communications of the ACM, 60(6), 84-90.
[3] Chen, Z. M., Wei, X. S., Wang, P., & Guo, Y. (2019). Multi-label image recognition with graph convolutional networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 5177-5186).
[4] Li, Y. (2022, January). Research and application of deep learning in image recognition. In 2022 IEEE 2nd international conference on power, electronics and computer applications (ICPECA) (pp. 994-999). IEEE.
[5] Li, G., Zhang, M., Li, J., Lv, F., & Tong, G. (2021). Efficient densely connected convolutional neural networks. Pattern Recognition, 109, 107610.
[6] Liu, Y., Lee, J., Park, M., Kim, S., Yang, E., Hwang, S. J., & Yang, Y. (2018). Learning to propagate labels: Transductive propagation network for few-shot learning. arXiv preprint arXiv:1805.10002.
[7] Vinyals, O., Blundell, C., Lillicrap, T., & Wierstra, D. (2016). Matching networks for one shot learning. Advances in Neural Information Processing Systems, 29.
[8] Finn, C., Abbeel, P., & Levine, S. (2017, July). Model-agnostic meta-learning for fast adaptation of deep networks. In International conference on machine learning (pp. 1126-1135). PMLR.
[9] Snell, J., Swersky, K., & Zemel, R. (2017). Prototypical networks for few-shot learning. Advances in Neural Information Processing Systems, 30.
[10] Ding, Y., Tian, X., Yin, L., Chen, X., Liu, S., Yang, B., & Zheng, W. (2019, September). Multi-scale relation network for few-shot learning based on meta-learning. In International Conference on Computer Vision Systems (pp. 343-352). Cham: Springer International Publishing.
[11] Fallah, A., Mokhtari, A., & Ozdaglar, A. (2020, June). On the convergence theory of gradient-based model-agnostic meta-learning algorithms. In International Conference on Artificial Intelligence and Statistics (pp. 1082-1092). PMLR.
[12] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., ... & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30.
[13] Ren, M., Triantafillou, E., Ravi, S., Snell, J., Swersky, K., Tenenbaum, J. B., ... & Zemel, R. S. (2018). Meta-learning for semi-supervised few-shot classification. arXiv preprint arXiv:1803.00676.
[14] Oreshkin, B., Rodríguez López, P., & Lacoste, A. (2018). Tadam: Task dependent adaptive metric for improved few-shot learning. Advances in Neural Information Processing Systems, 31.
[15] Wah, C., Branson, S., Welinder, P., Perona, P., & Belongie, S. (2011). The caltech-ucsd birds-200-2011 dataset.
[16] ZhuЃ, X., & GhahramaniЃн, Z. (2002). Learning from labeled and unlabeled data with label propagation.
[17] Kingma, D. P., & Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980.
How to cite this paper
Graph-based Embedding Propagation for Transductive Few-shot Classification
How to cite this paper: Kangwen Niu. (2024) Graph-based Embedding Propagation for Transductive Few-shot Classification. Advances in Computer and Communication, 5(4), 210-217.
DOI: https://dx.doi.org/10.26855/acc.2024.10.002
More Articles
- Research on Intelligent Batch Evidence Collection for Video Duration
- Research on a Low-code Data Analysis and Modeling Platform for Public Safety
- The Research on End-to-end Stock Recommendation Algorithm Based on Time-frequency Consistency
- Research on High-speed Digital Circuit Signal Transmission and Noise Suppression
- Industry 4.0: Manufacturing Informatics-driven Innovation
- ChatGPT Class Generative Artificial Intelligence Criminal Subject Problem Thinking