Filters








7 Hits in 2.6 sec

not-MIWAE: Deep Generative Modelling with Missing not at Random Data [article]

Niels Bruun Ipsen, Pierre-Alexandre Mattei, Jes Frellsen
2021 arXiv   pre-print
Specifically, a deep neural network enables us to flexibly model the conditional distribution of the missingness pattern given the data.  ...  We present an approach for building and fitting deep latent variable models (DLVMs) in cases where the missing process is dependent on the missing data.  ...  Again, such a setting can appear when there is strong geometric structure in the data (e.g. with images or proteins).  ... 
arXiv:2006.12871v2 fatcat:jsk4zkr45rgk3hz7mxtipf5ubi

Deep Generative Pattern-Set Mixture Models for Nonignorable Missingness [article]

Sahra Ghalebikesabi, Rob Cornish, Luke J. Kelly, Chris Holmes
2021 arXiv   pre-print
Our model explicitly learns to cluster the missing data into missingness pattern sets based on the observed data and missingness masks.  ...  Our model outperforms many common imputation algorithms, especially when the amount of missing data is high and the missingness mechanism is nonignorable.  ...  CH is supported by The Alan Turing Institute, Health Data Research UK, the Medical Research Council UK, the EPSRC through the Bayes4Health programme Grant EP/R018561/1, and AI for Science and Government  ... 
arXiv:2103.03532v1 fatcat:wfgghu6rxbbxfka75pxo36t3om

Identifiable Generative Models for Missing Not at Random Data Imputation [article]

Chao Ma, Cheng Zhang
2021 arXiv   pre-print
This is known as missing not at random (MNAR) data.  ...  Real-world datasets often have missing values associated with complex generative processes, where the cause of the missingness may not be fully observed.  ...  model for missing not at random).  ... 
arXiv:2110.14708v1 fatcat:rfsevenvprcirpw6mwicqkpy4q

Spatiotemporal Tensor Completion for Improved Urban Traffic Imputation

Ahmed Ben Said, Abdelkarim Erradi
2021 IEEE transactions on intelligent transportation systems (Print)  
We conduct comprehensive comparative study with two evaluation scenarios. In the first one, we simulate random missing values.  ...  In the second scenario, we simulate missing values at a given area and time duration.  ...  In [29] , the authors presented not-MIWAE, the not-missing-at-random IWAE to deal with data missing not at random.  ... 
doi:10.1109/tits.2021.3062999 fatcat:kl7r4ynhercfhgonu5ao6x7xt4

FragmGAN: Generative Adversarial Nets for Fragmentary Data Imputation and Prediction [article]

Fang Fang, Shenliao Bao
2022 arXiv   pre-print
for imputation with data Missing At Random (MAR) while no hint mechanism is needed.  ...  Unlike most of the other generative model based imputation methods that either have no theoretical guarantee or only consider Missing Completed At Random (MCAR), the proposed FragmGAN has theoretical guarantees  ...  ., 2018) first uses a Generative Adversarial Net (GAN) to impute data Missing Completed At Random (MCAR), which means the missingness occurs entirely at random without depending on any of the variables  ... 
arXiv:2203.04692v1 fatcat:qn5usfwx5zfzxia4g2ue7e7jem

Differentiable and Scalable Generative Adversarial Models for Data Imputation [article]

Yangyang Wu and Jun Wang and Xiaoye Miao and Wenjia Wang and Jianwei Yin
2022 arXiv   pre-print
Data imputation has been extensively explored to solve the missing data problem.  ...  DIM leverages a new masking Sinkhorn divergence function to make an arbitrary generative adversarial imputation model differentiable, while for such a differentiable imputation model, SSE can estimate  ...  data importance-weighted autoencoder) [9] , and not-MIWAE (not-missing-at-random importance-weighted autoencoder) [13] , and iii) GAN-based ones, such as GINN (graph imputation neural network) [15]  ... 
arXiv:2201.03202v1 fatcat:nhzoo7hixjha5opg5qb6kn2sb4

Reconstruction of Incomplete Wildfire Data using Deep Generative Models [article]

Tomislav Ivek, Domagoj Vlah
2022
Our deep latent variable generative model requires little to no feature engineering and does not necessarily rely on the specifics of scoring in the Data Challenge.  ...  The presented approach is not domain-specific and is amenable to application in other missing data recovery tasks with tabular or image-like information conditioned on auxiliary information.  ...  We thank Stjepan Šebek and Josip Žubrinić for valuable discussions and help in data preparation.  ... 
doi:10.48550/arxiv.2201.06153 fatcat:rlw67kzf3rhobmcsxm4qglgxpu