進階搜尋


   電子論文尚未授權公開,紙本請查館藏目錄
(※如查詢不到或館藏狀況顯示「閉架不公開」,表示該本論文不在書庫,無法取用。)
系統識別號 U0026-2408202017014600
論文名稱(中文) 具零編碼描述的虛擬智慧演化
論文名稱(英文) P-Learning: Virtual Intelligence Evolution with Zero Pattern of Class Embedding
校院名稱 成功大學
系所名稱(中) 資訊工程學系
系所名稱(英) Institute of Computer Science and Information Engineering
學年度 108
學期 2
出版年 109
研究生(中文) 龎在偉
研究生(英文) Tzai-Wei Pang
學號 P76071187
學位類別 碩士
語文別 英文
論文頁數 43頁
口試委員 指導教授-鄭憲宗
口試委員-許智威
口試委員-施嘉興
口試委員-陳盈鈞
中文關鍵字 零樣本學習  變分自編碼器  知識表示 
英文關鍵字 Zero-shot learning  Variational Auto-Encoder  Knowledge representation 
學科別分類
中文摘要 近年來零樣本學習是個重大且需要解決的問題,然而零樣本學習方法受限於需要不可見類描述的缺點。基於從無到有的演化概念,我們提出一個類似想法的機制用來生成不可見類的資訊。變分自編碼器是目前在深度學習中用來學習複雜資料的分布與表示的最好方法之一。本篇論文使用了以條件變分自編碼器分辨零樣本學習中可見類與不可見類。在缺少不可見類的類別編碼的情況下,學習可見類的分布後得到類別編碼的編碼器和一個可以概略得分類資料的分類器。以生成後條件變分自編碼器的類別隱表示作為可見類編碼以及計算得到不可見類的編碼。之後再得到給定的類別編碼與產生的類別編碼之間的對應。最後根據上述的描述,我們的建立一個類似樣本學習方法的分類器。我們使用植物疾病與其他經常應用於零樣本學習的資料集,我們展示我們的方法可以達到與普通零樣本學習相似的效能與在時間消耗上的優勢。
英文摘要 Zero-shot learning (ZSL) is a significant and hence to be developed problem. However, ZSL suffers a defect that the need of unseen class embedding. By the concept of evolution, we propose a mechanism that produces unseen class information is like an evolutionary procedure which is from none to one. Variational Auto-Encoder (VAE) is one of the state-of-the-art methods in deep learning to learn distribution and representation of complex data. In this work, we use Conditional Variational Autoencoder (CAVE) to distinguish the class data. In lacking unseen class embedding case, CVAE learns seen class distribution for class encoder and a classifier to roughly classify data. It notes that synthesizes latent representation as seen class embedding and calculated unseen class embedding. And then we align distribution between given class embedding and learned class embedding. Finally, based on these processes, our method constructs a classifier which is like general ZSL procedure. Experimental results on plant diseases dataset and others. We demonstrate that P-Learning can meet similar performance with ordinary ZSL work and has advantage in time consuming.
論文目次 摘要 2
Abstract 3
ACKNOWLEDGMENT 4
TABLE OF CONTENTS 5
LIST OF FIGURES 7
LIST OF TABLES 8
Chapter 1. Introduction and Motivation 9
Chapter 2. Background and Related work 14
2.1 Zero-shot learning 14
2.2 Generalized zero-shot learning 16
2.3 Variational Auto-Encoder 17
2.4 Conditional Variational Autoencoder 18
Chapter 3. Approach 19
3.1 Problem Description 19
3.2 Modeling VAE for input distribution 24
3.3 Aligning class embedding relation 26
3.4 Filtering data from input 28
Chapter 4. Implementation and Experiments 30
4.1 Implementation and Environment 30
4.1.1 Implementation of P-learning 30
4.1.2 Evaluating benchmark 32
4.1.3 Baseline methods and Environment 32
4.2 Experiments and Criterion 33
Chapter 5. Conclusions and Future Work 39
5.1 Conclusions 39
5.2 Future Work 40
Reference 41
參考文獻 [1] Kim, B.; Shin, S.; Jung, H. Variational Autoencoder-Based Multiple Image Captioning Using a Caption Attention Map. Appl. Sci. 2019, 9, 2699.
[2] Adrian Alan Pol ; Victor Berger ; Cecile Germain ; Gianluca Cerminara ; Maurizio Pierini, "Anomaly Detection with Conditional Variational Autoencoders" IEEE International Conference On Machine Learning And Applications (ICMLA), 2019
[3] Jian Ni, Shanghang Zhang, Haiyong Xie, "Dual Adversarial Semantics-Consistent Network for Generalized Zero-Shot Learning." IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2019
[4] Hyeonwoo Yu, Beomhee Lee. Zero-shot Learning via Simultaneous Generating and Learning. Conference on Neural Information Processing Systems (NeurIPS 2019)
[5] Yongqin Xian, Christoph H. Lampert, Bernt Schiele, Zeynep Akata. Zero-Shot Learning—A Comprehensive Evaluation of the Good, the Bad and the Ugly. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 41, NO. 9, SEPTEMBER 2019.
[6] Jeffrey Dalton, James Allan, and Pranav Mirajkar. Zero-shot video retrieval using content and concepts. 22nd ACM international conference on Information &Knowledge Management, pages 1857–1860. ACM, 2013.
[7] Yuming Shen, Li Liu, Fumin Shen, and Ling Shao. Zero-shot sketch-image hashing. arXiv preprint arXiv:1803.02284,2018.
[8] Ziming Zhang and Venkatesh Saligrama. Zero-shot learning via semantic similarity embedding. The IEEE International Conference on Computer Vision (ICCV),pages 4166–4174, 2015
[9] Christoph H Lampert, Hannes Nickisch, and Stefan Harmeling. Attribute-based classification for zero-shot visual object categorization. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014.
[10] Xiaodong Yu and Yiannis Aloimonos. Attribute-based transfer learning for object categorization with zero/one training example. European Conf. on Computer Vision, 2010.
[11] Devraj Mandal, Sanath Narayan, Saikumar Dwivedi, Vikram Gupta, Shuaib Ahmed, Fahad Shahbaz Khan, and Ling Shao. Out-of-Distribution Detection for Generalized Zero-Shot Action Recognition. IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2019
[12] G. Shalev, Y. Adi, and J. Keshet. Out-of-distribution detection using multiple semantic label representations. InNIPS,2018
[13] Arash Vahdat, Jan Kautz. NVAE: A Deep Hierarchical Variational Autoencoder. IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2020.
[14] Edgar Schönfeld, Sayna Ebrahimi, Samarth Sinha, Trevor Darrell, Zeynep Akata, " Generalized Zero- and Few-Shot Learning via Aligned Variational Autoencoders," IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2019
[15] Yang Liu, Jishun Guo, Deng Cai, Xiaofei He. "Attribute Attention for Semantic Disambiguation in Zero-Shot Learning." ICCV (2019).
[16] Catherine Wah, Steve Branson, Peter Welinder, Pietro Per-ona, and Serge Belongie. The caltech-ucsd birds-200-2011dataset. 2011
[17] Genevieve Patterson and James Hays. Sun attribute database: Discovering, annotating, and recognizing scene attributes. In 2012 IEEE Conference on Computer Vision and Pattern Recognition, pages 2751–2758. IEEE, 2012.
[18] Y. Xian, C. H. Lampert, B. Schiele, and Z. Akata. Zero-shot learning-a comprehensive evaluation of the good, the bad and the ugly. IEEE transactions on pattern analysis and machine intelligence, 2018
[19] Samir Bhattarai. New Plant Diseases Dataset. Kaggle.
[20] Andrea Frome*, Greg S. Corrado*, Jonathon Shlens*, Samy BengioJeffrey Dean, Marc’Aurelio Ranzato, Tomas Mikolov. DeViSE: A Deep Visual-Semantic Embedding Model. NIPS 2013.
[21] Yongqin Xian, Tobias Lorenz, Bernt Schiele, Zeynep Akata. Feature Generating Networks for Zero-Shot Learning. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018.
[22] Soravit Changpinyo, Wei-Lun Chao, Boqing Gong, and FeiSha. Synthesized classifiers for zero-shot learning. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016
[23] Christoph H. Lampert, Hannes Nickisch, Stefan Harmeling. Learning To Detect Unseen Object Classes by Between-Class Attribute Transfer. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2009.
[24] Zhengming Ding, Ming Shao, Yun Fu. Generative Zero-Shot Learning via Low-Rank Embedded Semantic Dictionary. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 41, NO. 12, DECEMBER 2019.
論文全文使用權限
  • 同意授權校內瀏覽/列印電子全文服務,於2025-09-01起公開。
  • 同意授權校外瀏覽/列印電子全文服務,於2025-09-01起公開。


  • 如您有疑問,請聯絡圖書館
    聯絡電話:(06)2757575#65773
    聯絡E-mail:etds@email.ncku.edu.tw