基于特征變換和度量網(wǎng)絡(luò )的小樣本學(xué)習算法

王多瑞; 杜楊; 董蘭芳; 胡衛明; 李兵

doi:10.16383/j.aas.c210903

基于特征變換和度量網(wǎng)絡(luò )的小樣本學(xué)習算法

doi: 10.16383/j.aas.c210903

王多瑞^{1, 2,},
杜楊^{2, 3},
董蘭芳^1,,
胡衛明^2,,
李兵²

1.
中國科學(xué)技術(shù)大學(xué) 合肥 230026
2.
中國科學(xué)院自動(dòng)化研究所模式識別國家重點(diǎn)實(shí)驗室北京100190
3.
阿里巴巴科技(北京)有限公司北京 100016

基金項目: 國家重點(diǎn)研發(fā)計劃(2018AAA0102802), 國家自然科學(xué)基金(62036011, 62192782, 61721004), 中國科學(xué)院前沿科學(xué)重點(diǎn)研究計劃(QYZDJ-SSW-JSC040)資助

詳細信息

作者簡(jiǎn)介:
王多瑞：2021年獲得中國科學(xué)技術(shù)大學(xué)碩士學(xué)位. 主要研究方向為小樣本學(xué)習, 目標檢測.E-mail: wangduor@mail.ustc.edu.cn

杜楊：2019年獲得中國科學(xué)院自動(dòng)化研究所博士學(xué)位. 主要研究方向為行為識別, 醫學(xué)圖像處理.E-mail: jingzhou.dy@alibaba-inc.com

董蘭芳：中國科學(xué)技術(shù)大學(xué)副教授. 1994年獲得中國科學(xué)技術(shù)大學(xué)碩士學(xué)位. 主要研究方向為圖像與視頻智能分析, 知識圖譜與對話(huà)系統, 數值模擬與三維重建.E-mail: lfdong@ustc.edu.cn

胡衛明：中國科學(xué)院自動(dòng)化研究所研究員. 1998年獲得浙江大學(xué)博士學(xué)位. 主要研究方向為視覺(jué)運動(dòng)分析, 網(wǎng)絡(luò )不良信息識別和網(wǎng)絡(luò )入侵檢測. 本文通信作者.E-mail: wmhu@nlpr.ia.ac.cn

李兵：中國科學(xué)院自動(dòng)化研究所研究員. 2009年獲得北京交通大學(xué)博士學(xué)位. 主要研究方向為網(wǎng)絡(luò )內容安全, 智能圖像信號處理.E-mail: bing.li@ia.ac.cn

計量
- 文章訪(fǎng)問(wèn)數: 639
- HTML全文瀏覽量: 194
- PDF下載量: 175
- 被引次數: 0
出版歷程
- 收稿日期: 2021-09-20
- 錄用日期: 2021-12-11
- 網(wǎng)絡(luò )出版日期: 2023-09-11
- 刊出日期: 2024-07-23

Feature Transformation and Metric Networks for Few-shot Learning

WANG Duo-Rui^{1, 2
,},
DU Yang^{2, 3},
DONG Lan-Fang^1
,,
HU Wei-Ming^2
,,
LI Bing²

1.
University of Science and Technology of China, Hefei 230026
2.
National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing 100190
3.
Alibaba (Beijing) Group, Beijing 100016

Funds: Supported by National Key Research and Development Program of China (2018AAA0102802), National Natural Science Foundation of China (62036011, 62192782, 61721004), and Key Research Program of Frontier Sciences of Chinese Academy of Sciences (QYZDJ-SSW-JSC040)

More Information

Author Bio:
WANG Duo-Rui　He received his master degree from University of Science and Technology of China in 2021. His research interest covers few-shot learning and object detection

DU Yang　He received his Ph.D. degree from the Institute of Automation, Chinese Academy of Sciences. His research interest covers action recognition and medical image processing

DONG Lan-Fang　Associate professor at University of Science and Technology of China. She received her master degree from University of Science and Technology of China in 1994. Her research interest covers image and video intelligent analysis, knowledge mapping and dialogue systems, and numerical simulation and 3D reconstruction

HU Wei-Ming　Professor at the Institute of Automation, Chinese Academy of Sciences. He received his Ph.D. degree from Zhejiang University in 1998. His research interest covers visual motion analysis, recognition of web objectionable information, and network intrusion detection. Corresponding author of this paper

LI Bing　Professor at the Institute of Automation, Chinese Academy of Sciences. He received his Ph.D. degree from Beijing Jiaotong University in 2009. His research interest covers the web content security and intelligent image signal process

摘要

摘要: 在小樣本分類(lèi)任務(wù)中, 每個(gè)類(lèi)別可供訓練的樣本數量非常有限. 因此在特征空間中同類(lèi)樣本分布稀疏, 異類(lèi)樣本間邊界模糊. 提出一種新的基于特征變換和度量網(wǎng)絡(luò )(Feature transformation and metric networks, FTMN)的小樣本學(xué)習算法用于小樣本分類(lèi)任務(wù). 算法通過(guò)嵌入函數將樣本映射到特征空間, 并計算輸入該樣本與所屬類(lèi)別中心的特征殘差. 構造一個(gè)特征變換函數對該殘差進(jìn)行學(xué)習, 使特征空間內的樣本特征經(jīng)過(guò)該函數后向同類(lèi)樣本中心靠攏. 利用變換后的樣本特征更新類(lèi)別中心, 使各類(lèi)別中心間的距離增大. 算法進(jìn)一步構造了一種新的度量函數, 對樣本特征中每個(gè)局部特征點(diǎn)的度量距離進(jìn)行聯(lián)合表達, 該函數能夠同時(shí)對樣本特征間的夾角和歐氏距離進(jìn)行優(yōu)化. 算法在小樣本分類(lèi)任務(wù)常用數據集上的優(yōu)秀表現證明了算法的有效性和泛化性.
- 特征變換 /
- 度量學(xué)習 /
- 小樣本學(xué)習 /
- 殘差學(xué)習
Abstract: For few-shot classification, training samples for each class are highly limited. Consequently, samples from the same class tend to distribute sparsely while boundaries between different classes are indistinct in the feature space. Therefore, a novel few-shot learning algorithm based on feature transformation and metric networks (FTMN) is proposed for few-shot learning. The algorithm maps samples to the feature space through an embedding function and calculates the residual between the input features and their class center. A feature transformation function is then constructed to learn from the residual, enabling input features to move closer to their class center after transformation. The transformed features are used to update the class centers, increasing the distance between centers of different classes. Furthermore, the algorithm introduces a novel metric function that jointly expresses the metric distances of each point within the features. The metric function simultaneously optimizes both cosine similarity and Euclidean distance. The performance of the algorithm on commonly used datasets for few-shot classification validates its effectiveness and generalization ability.
- Feature transformation /
- metric learning /
- few-shot learning /
- residual learning

HTML全文

圖 1 特征變換和度量網(wǎng)絡(luò )模型

Fig. 1 Model of feature transformation and metric networks

下載: 全尺寸圖片幻燈片

圖 2 網(wǎng)絡(luò )中關(guān)鍵函數的結構

Fig. 2 Structure of important functions of networks

下載: 全尺寸圖片幻燈片

表 1 網(wǎng)絡(luò )模型的嵌入函數與重要結構

Table 1 Embedding function and important structures of networks

模型名稱(chēng)	嵌入函數	重要結構
MN	4層卷積網(wǎng)絡(luò )	注意力長(cháng)短時(shí)記憶網(wǎng)絡(luò )
ProtoNet^[12]	4層卷積網(wǎng)絡(luò )	“原型”概念、使用歐氏距離進(jìn)行度量
RN	4層卷積網(wǎng)絡(luò )	卷積神經(jīng)網(wǎng)絡(luò )作為度量函數
EGNN	4層卷積網(wǎng)絡(luò )	邊標簽預測節點(diǎn)類(lèi)別
EGNN + Transduction^[22]	ResNet-12	邊標簽預測節點(diǎn)類(lèi)別、轉導和標簽傳遞
DN4^[24]	ResNet-12	局部描述子、圖像與類(lèi)別間的相似性度量
DC^[25]	4層卷積網(wǎng)絡(luò )	稠密分類(lèi)
DC + IMP^[25]	4層卷積網(wǎng)絡(luò )	稠密分類(lèi)、神經(jīng)網(wǎng)絡(luò )遷移
FTMN	4層卷積網(wǎng)絡(luò )	特征變換模塊、特征度量模塊
FTMN-R12	ResNet-12	特征變換模塊、特征度量模塊

下載: 導出CSV

表 2 在Omniglot數據集上的小樣本分類(lèi)性能(%)

Table 2 Few-shot classification performance on Omniglot dataset (%)

模型	5-類(lèi)		20-類(lèi)
模型	1-樣本	5-樣本	1-樣本	5-樣本
MN	98.1	98.9	93.8	98.5
ProtoNet^[12]	98.8	99.7	96.0	98.9
SN	97.3	98.4	88.2	97.0
RN	99.6 ± 0.2	99.8 ± 0.1	97.6 ± 0.2	99.1 ± 0.1
SM^[15]	98.4	99.6	95.0	98.6
MetaNet^[16]	98.95	—	97.00	—
MANN^[17]	82.8	94.9	—	—
MAML^[18]	98.7 ± 0.4	99.9 ± 0.1	95.8 ± 0.3	98.9 ± 0.2
MMNet^[26]	99.28 ± 0.08	99.77 ± 0.04	97.16 ± 0.10	98.93 ± 0.05
FTMN	99.7 ± 0.1	99.9 ± 0.1	98.3 ± 0.1	99.5 ± 0.1

下載: 導出CSV

表 3 在miniImageNet數據集上的小樣本分類(lèi)性能 (%)

Table 3 Few-shot classification performance on miniImageNet dataset (%)

模型	5-類(lèi)
模型	1-樣本	5-樣本
MN	43.40 ± 0.78	51.09 ± 0.71
ML-LSTM^[11]	43.56 ± 0.84	55.31 ± 0.73
ProtoNet^[12]	49.42 ± 0.78	68.20 ± 0.66
RN	50.44 ± 0.82	65.32 ± 0.70
MetaNet^[16]	49.21 ± 0.96	—
MAML^[18]	48.70 ± 1.84	63.11 ± 0.92
EGNN	—	66.85
EGNN + Transduction^[22]	—	76.37
DN4^[24]	51.24 ± 0.74	71.02 ± 0.64
DC^[25]	62.53 ± 0.19	78.95 ± 0.13
DC + IMP^[25]	—	79.77 ± 0.19
MMNet^[26]	53.37 ± 0.08	66.97 ± 0.09
PredictNet^[27]	54.53 ± 0.40	67.87 ± 0.20
DynamicNet^[28]	56.20 ± 0.86	72.81 ± 0.62
MN-FCE^[29]	43.44 ± 0.77	60.60 ± 0.71
MetaOptNet^[30]	60.64 ± 0.61	78.63 ± 0.46
FTMN	59.86 ± 0.91	75.96 ± 0.82
FTMN-R12	61.33 ± 0.21	79.59 ± 0.47

下載: 導出CSV

表 4 在CUB-200、CIFAR-FS和tieredImageNet數據集上的小樣本分類(lèi)性能(%)

Table 4 Few-shot classification performance on CUB-200, CIFAR-FS and tieredImageNet datasets (%)

模型	CUB-200 5-類(lèi)		CIFAR-FS 5-類(lèi)		tieredImageNet 5-類(lèi)
模型	1-樣本	5-樣本	1-樣本	5-樣本	1-樣本	5-樣本
MN	61.16 ± 0.89	72.86 ± 0.70	—	—	—	—
ProtoNet^[12]	51.31 ± 0.91	70.77 ± 0.69	55.5 ± 0.7	72.0 ± 0.6	53.31 ± 0.89	72.69 ± 0.74
RN	62.45 ± 0.98	76.11 ± 0.69	55.0 ± 1.0	69.3 ± 0.8	54.48 ± 0.93	71.32 ± 0.78
MAML^[18]	55.92 ± 0.95	72.09 ± 0.76	58.9 ± 1.9	71.5 ± 1.0	51.67 ± 1.81	70.30 ± 1.75
EGNN	—	—	—	—	63.52 ± 0.52	80.24 ± 0.49
DN4^[24]	53.15 ± 0.84	81.90 ± 0.60	—	—	—	—
MetaOptNet^[30]	—	—	72.0 ± 0.7	84.2 ± 0.5	65.99 ± 0.72	81.56 ± 0.53
FTMN-R12	69.58 ± 0.36	85.46 ± 0.79	70.3 ± 0.5	82.6 ± 0.3	62.14 ± 0.63	81.74 ± 0.33

下載: 導出CSV

表 5 消融實(shí)驗結果 (%)

Table 5 Results of ablation study (%)

模型	5-類(lèi)
模型	1-樣本	5-樣本
ProtoNet-4C	49.42 ± 0.78	68.20 ± 0.66
ProtoNet-8C	51.18 ± 0.73	70.23 ± 0.46
ProtoNet-Trans-4C	53.47 ± 0.46	71.33 ± 0.23
ProtoNet-M-4C	56.54 ± 0.57	73.46 ± 0.53
ProtoNet-VLAD-4C	52.46 ± 0.67	70.83 ± 0.62
Trans*-M-4C	59.86 ± 0.91	67.86 ± 0.56
僅使用余弦相似度	54.62 ± 0.57	72.58 ± 0.38
僅使用歐氏距離	55.66 ± 0.67	73.34 ± 0.74
FTMN	59.86 ± 0.91	75.96 ± 0.82

下載: 導出CSV

<button id="qm3rj"><thead id="qm3rj"></thead></button>

亚洲第一网址_国产国产人精品视频69_久久久久精品视频_国产精品第九页

參考文獻(30)

[1]	Szegedy C, Liu W, Jia Y Q, Sermanet P, Reed S, Anguelov D, et al. Going deeper with convolutions. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston, USA: IEEE, 2015. 1?9
[2]	He K M, Zhang X Y, Ren S Q, Sun J. Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE, 2016. 770?778
[3]	Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks. In: Proceedings of the 26th International Conference on Neural Information Processing Systems. Lake Tahoe, USA: NIPS, 2012. 1106?1114
[4]	Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. In: Proceedings of the 3rd International Conference on Learning Representations. San Diego, USA: ICLR, 2015.
[5]	劉穎, 雷研博, 范九倫, 王富平, 公衍超, 田奇. 基于小樣本學(xué)習的圖像分類(lèi)技術(shù)綜述. 自動(dòng)化學(xué)報, 2021, 47(2): 297?315 doi: 10.16383/j.aas.c190720 Liu Ying, Lei Yan-Bo, Fan Jiu-Lun, Wang Fu-Ping, Gong Yan-Chao, Tian Qi. Survey on image classification technology based on small sample learning. Acta Automatica Sinica, 2021, 47(2): 297?315 doi: 10.16383/j.aas.c190720
[6]	Miller E G, Matsakis N E, Viola P A. Learning from one example through shared densities on transforms. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Hilton Head Island, USA: IEEE, 2000. 464?471
[7]	Li F F, Fergus R, Perona P. One-shot learning of object categories. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2006, 28(4): 594?611
[8]	Lake B M, Salakhutdinov R, Gross J, Tenenbaum J B. One shot learning of simple visual concepts. In: Proceedings of the 33rd Annual Meeting of the Cognitive Science Society. Boston, USA: CogSci, 2011. 2568?2573
[9]	Lake B M, Salakhutdinov R, Tenenbaum J B. Human-level concept learning through probabilistic program induction. Science, 2015, 350(11): 1332?1338
[10]	Edwards H, Storkey A J. Towards a neural statistician. In: Proceedings of the 5th International Conference on Learning Representations. Toulon, France: ICLR, 2017.
[11]	Vinyals O, Blundell C, Lillicrap T, Kavukcuoglu K, Wierstra D. Matching networks for one shot learning. In: Proceedings of the 30th International Conference on Neural Information Processing Systems. Barcelona, Spain: 2016. 3637?3645
[12]	Snell J, Swersky K, Zemel R. Prototypical networks for few-shot learning. In: Proceedings of the 31th International Conference on Neural Information Processing Systems. Long Beach, USA: 2017. 4080?4090
[13]	Koch G, Zemel R, Salakhutdinov R. Siamese neural networks for one-shot image recognition. In: Proceedings of the 32nd International Conference on Machine Learning. Lille, France: JMLR, 2015.
[14]	Sung F, Yang Y X, Zhang L, Xiang T, Torr P H S, Hospedales T M. Learning to compare: Relation network for few-shot learning. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018. 1199?1208
[15]	Kaiser L, Nachum O, Roy A, Bengio S. Learning to remember rare events. In: Proceedings of the 5th International Conference on Learning Representations. Toulon, France: ICLR, 2017.
[16]	Munkhdalai T, Yu H. Meta networks. In: Proceedings of the 34th International Conference on Machine Learning. Sydney, Australia: JMLR.org, 2017. 2554?2563
[17]	Santoro A, Bartunov S, Botvinick M, Wierstra D, Lillicrap T. Meta-learning with memory-augmented neural networks. In: Proceedings of the 33rd International Conference on Machine Learning. New York, USA: PMLR, 2016. 1842?1850
[18]	Finn C, Abbeel P, Levine S. Model-agnostic meta-learning for fast adaptation of deep networks. In: Proceedings of the 34th International Conference on Machine Learning. Sydney, Australia: JMLR.org, 2017. 1126?1135
[19]	Arandjelovic R, Gronat P, Torii A, Pajdla T, Sivic J. Net-VLAD: CNN architecture for weakly supervised place recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, USA: IEEE, 2016. 5297?5307
[20]	Jégou H, Douze M, Schmid C, Pérez P. Aggregating local descriptors into a compact image representation. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. San Francisco, USA: IEEE, 2010. 3304?3311
[21]	Bertinetto L, Henriques J F, Torr P H, Vedaldi A. Meta-learning with differentiable closed-form solvers. In: Proceedings of the 7th International Conference on Learning Representations. New Orleans, USA: ICLR, 2019.
[22]	Kim J, Kim T, Kim S, Yoo C D. Edge-labeling graph neural network for few-shot learning. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, USA: IEEE, 2019. 11?20
[23]	Yue Z Q, Zhang H W, Sun Q R, Hua X S. Interventional few-shot learning. In: Proceedings of the 34th International Conference on Neural Information Processing Systems. Vancouver, Canada: Curran Associates Incorporated, 2020. Article No. 230
[24]	Li W B, Wang L, Xu J L, Huo J, Gao Y, Luo J B. Revisiting local descriptor based image-to-class measure for few-shot learning. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, USA: IEEE, 2019. 7253?7260
[25]	Lifchitz Y, Avrithis Y, Picard S, Bursuc A. Dense classification and implanting for few-shot learning. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, USA: IEEE, 2019. 9250?9259
[26]	Cai Q, Pan Y W, Yao T, Yan C G, Mei T. Memory matching networks for one-shot image recognition. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018. 4080?4088
[27]	Qiao S Y, Liu C X, Shen W, Yuille A L. Few-shot image recognition by predicting parameters from activations. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018. 7229?7238
[28]	Gidaris S, Komodakis N. Dynamic few-shot visual learning without forgetting. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, USA: IEEE, 2018. 4367?4375
[29]	Ravi S, Larochelle H. Optimization as a model for few-shot learning. In: Proceedings of the 5th International Conference on Learning Representations. Toulon, France: ICLR, 2017.
[30]	Lee K, Maji S, Ravichandran A, Soatto S. Meta-learning with differentiable convex optimization. In: Proceedings of the IEEE/ CVF Conference on Computer Vision and Pattern Recognition. Long Beach, USA: IEEE, 2019. 10649?10657