Processing of missing data by neural networks

Authors: Marek Śmieja, Łukasz Struski, Jacek Tabor, Bartosz Zieliński, Przemysław Spurek

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results performed on different types of architectures show that our method gives better results than typical imputation strategies and other methods dedicated for incomplete data.
Researcher Affiliation Academia Marek Smieja marek.smieja@uj.edu.pl Łukasz Struski lukasz.struski@uj.edu.pl Jacek Tabor jacek.tabor@uj.edu.pl Bartosz Zieli nski bartosz.zielinski@uj.edu.pl Przemysław Spurek przemyslaw.spurek@uj.edu.pl Faculty of Mathematics and Computer Science Jagiellonian University Łojasiewicza 6, 30-348 Kraków, Poland
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes The code implementing the proposed method is available at https://github.com/lstruski/ Processing-of-missing-data-by-neural-networks.
Open Datasets Yes As a data set, we used grayscale handwritten digits retrieved from MNIST database. For each image of the size 28 28 = 784 pixels, we removed a square patch of the size5 13 13. The location of the patch was uniformly sampled for each image. AE used in the experiments consists of 5 hidden layers with 256, 128, 64, 128, 256 neurons in subsequent layers. The first layer was parametrized by Re LU activation functions, while the remaining units used sigmoids6.
Dataset Splits Yes We applied double 5-fold cross-validation procedure to report classification results and we tuned required hyper-parameters.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments.
Software Dependencies No The paper mentions using certain software tools like 'mice' and 'gmm' and refers to implementations in 'R', but it does not provide specific version numbers for any software dependencies required to replicate the experiments.
Experiment Setup Yes AE used in the experiments consists of 5 hidden layers with 256, 128, 64, 128, 256 neurons in subsequent layers. The first layer was parametrized by Re LU activation functions, while the remaining units used sigmoids6. As describe in Section 1, our model assumes that there is no complete data in training phase. Therefore, the loss function was computed based only on pixels from outside the mask.