Learning from incomplete data with infinite imputations

Dick, Uwe; Haider, Peter; Scheffer, Tobias

doi:10.1145/1390156.1390186

Cited by 31 publications

(14 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Building upon this generative model strategy, some approaches have considered integrating out the missing values based on a simple logistic regression function [6], [7]. Other versions of this approach proposed an explicit simultaneous learning of the model and the decision function [8], [9]. While probabilistic generative models provided a nice and elegant approach to the incomplete-features problem showing good results on small datasets, they are not suitable for many modern machine learning applications because:…”

Section: A Related Workmentioning

confidence: 99%

Learning from Incomplete Features by Simultaneous Training of Neural Networks and Sparse Coding

Caiafa¹,

Wang

Solé-Casals

et al. 2021

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

In this paper, the problem of training a classifier on a dataset with incomplete features is addressed. We assume that different subsets of features (random or structured) are available at each data instance. This situation typically occurs in the applications when not all the features are collected for every data sample. A new supervised learning method is developed to train a general classifier, such as a logistic regression or a deep neural network, using only a subset of features per sample, while assuming sparse representations of data vectors on an unknown dictionary. Sufficient conditions are identified, such that, if it is possible to train a classifier on incomplete observations so that their reconstructions are well separated by a hyperplane, then the same classifier also correctly separates the original (unobserved) data samples. Extensive simulation results on synthetic and well-known datasets are presented that validate our theoretical findings and demonstrate the effectiveness of the proposed method compared to traditional data imputation approaches and one state-of-the-art algorithm.

show abstract

Section: A Related Workmentioning

confidence: 99%

Learning from Incomplete Features by Simultaneous Training of Neural Networks and Sparse Coding

Caiafa¹,

Wang

Solé-Casals

et al. 2021

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

show abstract

“…However, this method may miss some important information, and several works have demonstrated the dangers of simply removing cases using the case deletion [6,7]. The next one is to learn without handing of missing values such as learning Bayesian Networks [8], Artificial Neural Networks [9,10] and missing values fish-net learning algorithms [11].…”

Section: A Related Workmentioning

confidence: 99%

Examination of Reliability of Missing Value Recovery in Data Mining

Liu

Dai

2014

2014 IEEE International Conference on Data Mining Workshop

View full text Add to dashboard Cite

Missing data imputation is an important task in cases where it is crucial to use all available data and no discard records with missing values. However, most of the existing algorithms are focused on missing at random (MAR) or missing completely at random (MCAR). In this paper, an information decomposition imputation (IDIM) algorithm using fuzzy membership function is proposed for addressing the missing value problem which is not missing at random (NMAR). The reliability of missing value recovery at not missing at random is examined. Firstly, this paper will discuss the proposed IDIM algorithm with detailed examples. Then, the reliability of the proposed approach is evaluated with extensive experiments compared with some typical algorithms, and the results demonstrate that the proposed algorithm has a higher accuracy rate than the exiting imputation methods in terms of normal root mean square error (NRMSE) and predictive accuracy at different set of data with missing values, which shows our method is more reliable in imputing missing values.

show abstract

“…In the first approach, case deletion [8] is a method which is used basically to ignore those cases that come along with the missing values and to complete the learning progression only to utilize the residual instances. For the second technique of the learning without handling of missing data schemes, such as Bayesian Networks method, Artificial Neural Networks method, and some more approaches are demonstrated in [9], [10]. The third approach is entirely dissimilar to the above two techniques.…”

Section: II Related Workmentioning

confidence: 99%

An Estimation of Missing Values by Modified Mixed Kernels

2014

International Conference on Computer Science and Information Systems (ICSIS’2014) Oct 17-18, 2014 Dubai (UAE)

View full text Add to dashboard Cite

In statistical practices, difficulties of missing data are universal. Several techniques are used to handle this dilemma of missing data. They include both old approaches, which require only a small amount of mathematical computations and new approaches, which require additional difficult computations that are ever easier for social work researchers to carry out the statistical programming softwares. In the existing system, there is a novel setting of missing data imputation, i.e. imputing in mixed-attribute data sets. This system offers two consistent estimators for discrete and continuously missing target values, correspondingly. After that a mixture-kernel based iterative estimator is offered to impute mixed-attribute data sets. In this method, the local kernel and global kernel are used and linear combination of these mixed kernels is used. Nevertheless, the accuracy of the system is decreased with the large number of data samples. Unquestionably it will degrade the performance of the system. To improve the performance and to increase the accuracy of the system we proposed three approaches. First we introduce the local kernal RBF using KL divergence, secondly we introduce the global kernal polynomial using probability distribution and finally mixed kernels in piece level combination instead of linear combination. From the experimental result we can obtain that the proposed system is much more effective than the existing system. The performance also is shown to have improved in this proposed system.

show abstract

Learning from incomplete data with infinite imputations

Cited by 31 publications

References 6 publications

Learning from Incomplete Features by Simultaneous Training of Neural Networks and Sparse Coding

Learning from Incomplete Features by Simultaneous Training of Neural Networks and Sparse Coding

Examination of Reliability of Missing Value Recovery in Data Mining

An Estimation of Missing Values by Modified Mixed Kernels

Contact Info

Product

Resources

About