MisConv: Convolutional Neural Networks for Missing Data

被引:2
|
作者
Likowski, Marcin Przewiez [1 ]
Smieja, Marek [1 ]
Struski, Lukasz [1 ]
Tabor, Jacek [1 ]
机构
[1] Jagiellonian Univ, Fac Math & Comp Sci, 6 Lojasiewicza St, PL-30348 Krakow, Poland
关键词
CHAINED EQUATIONS; IMPUTATION; MACHINE; VALUES;
D O I
10.1109/WACV51458.2022.00297
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Processing of missing data by modern neural networks, such as CNNs, remains a fundamental, yet unsolved challenge, which naturally arises in many practical applications, like image inpainting or autonomous vehicles and robots. While imputation-based techniques are still one of the most popular solutions, they frequently introduce unreliable information to the data and do not take into account the uncertainty of estimation, which may be destructive for a machine learning model. In this paper, we present MisConv, a general mechanism, for adapting various CNN architectures to process incomplete images. By modeling the distribution of missing values by the Mixture of Factor Analyzers, we cover the spectrum of possible replacements and find an analytical formula for the expected value of convolution operator applied to the incomplete image. The whole framework is realized by matrix operations, which makes MisConv extremely efficient in practice. Experiments performed on various image processing tasks demonstrate that MisConv achieves superior or comparable performance to the state-of-the-art methods.
引用
收藏
页码:2917 / 2926
页数:10
相关论文
共 50 条
  • [1] Geographical Imputation of Missing Poaceae Pollen Data via Convolutional Neural Networks
    Navares, Ricardo
    Luis Aznarte, Jose
    [J]. ATMOSPHERE, 2019, 10 (11)
  • [2] URBAN LAND COVER CLASSIFICATION WITH MISSING DATA USING DEEP CONVOLUTIONAL NEURAL NETWORKS
    Kampffmeyer, Michael
    Salberg, Arnt-Borre
    Jenssen, Robert
    [J]. 2017 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM (IGARSS), 2017, : 5161 - 5164
  • [3] Convolutional Neural Networks in Predicting Missing Text in Arabic
    Souri, Adnan
    Alachhab, Mohamed
    Eddine Elmohajir, Badr
    Zbakh, Abdelali
    [J]. INTERNATIONAL JOURNAL OF ADVANCED COMPUTER SCIENCE AND APPLICATIONS, 2019, 10 (06) : 520 - 527
  • [4] Processing of missing data by neural networks
    Smieja, Marek
    Struski, Lukasz
    Tabor, Jacek
    Zielinski, Bartosz
    Spurek, Przemyslaw
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 31 (NIPS 2018), 2018, 31
  • [5] Urban Land Cover Classification With Missing Data Modalities Using Deep Convolutional Neural Networks
    Kampffmeyer, Michael
    Salberg, Arnt-Borre
    Jenssen, Robert
    [J]. IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, 2018, 11 (06) : 1758 - 1768
  • [6] CLASSIFICATION BASED ON MISSING FEATURES IN DEEP CONVOLUTIONAL NEURAL NETWORKS
    Milosevic, N.
    Rackovic, M.
    [J]. NEURAL NETWORK WORLD, 2019, 29 (04) : 221 - 234
  • [7] Recurrent neural networks for missing or asynchronous data
    Bengio, Y
    Gingras, F
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 8: PROCEEDINGS OF THE 1995 CONFERENCE, 1996, 8 : 395 - 401
  • [8] Analysis of missing data with artificial neural networks
    Pastor, JBN
    Vidal, JML
    [J]. PSICOTHEMA, 2000, 12 (03) : 503 - 510
  • [9] Handling missing data through deep convolutional neural network
    Khan, Hufsa
    Wang, Xizhao
    Liu, Han
    [J]. INFORMATION SCIENCES, 2022, 595 : 278 - 293
  • [10] Imputation of missing data with neural networks for classification
    Choudhury, Suyra Jyoti
    Pal, Nikhil R.
    [J]. KNOWLEDGE-BASED SYSTEMS, 2019, 182