Swamping and masking in Markov boundary discovery

Liu, Xuqing; Liu, Xinsheng

doi:10.1007/s10994-016-5545-0

Cited by 14 publications

(10 citation statements)

References 20 publications

(39 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When the faithfulness assumption is violated, the MB of learnt from the data may not be unique (Pena et al, 2007;Statnikov et al, 2013). To deal with the violation of the faithfulness assumption, some research work has been done for identifying multiple MBs without the assumption, such as KIAMB (Pena et al, 2007), TIE* (Statnikov et al, 2013), SGAI (Yu et al, 2017), LCMB (Liu and Liu, 2016), and WLCMB (Liu and Liu, 2016). KIAMB was the first attempt to learn multiple MBs, but it needs to run multiple times and cannot guarantee finding all possible MBs of the class variable.…”

Section: Overview Of Constraint-based Methodsmentioning

confidence: 99%

“…In this subsection, we will discuss six representative MB learning algorithms for tackling the situation where the faithfulness or causal sufficiency assumption is violated, i.e., KI-AMB (Pena et al, 2007), TIE* (Statnikov et al, 2013), SGAI (Yu et al, 2017), LCMB (Liu and Liu, 2016), WLCMB (Liu and Liu, 2016), and M3B (Yu et al, 2018c).…”

Section: Methods Of Mb Learning With Relaxed Assumptionsmentioning

confidence: 99%

“…Existing causality-based feature selection algorithms adopt either the SFBS or the IFBS strategy, and they employ a selection criterion Φ, such as information gain, independence tests, and score criteria, to add/remove features to/from S. until no features in S are not removed GSMB (Margaritis and Thrun, 2000) IAMB IAMBnPC (Tsamardinos et al, 2003b) IAMB-IP Fast-IAMB (Yaramakala and Margaritis, 2005) Inter-IAMB (Tsamardinos et al, 2003b) Inter-IAMBnPC (Tsamardinos et al, 2003b (Ling et al, 2019) EEMB (Wang et al, 2020) MB learning with relaxed assumptions (e.g. the faithfulness assumption or causal sufficiency assumption) KIAMB (Pena et al, 2007) TIE* (Statnikov et al, 2013) SGAI (Yu et al, 2017) LCMB (Liu and Liu, 2016) WLCMB (Liu and Liu, 2016) M3B (Yu et al, 2018c) MB learning with special purpose (e.g. multiple data sets, distribution shift, and weak supervision) MIMB (Yu et al, 2018a) MCFS MIAMB and MKIAMB BASSUM (Cai et al, 2011) Semi-IAMB (Sechidis and Brown, 2018)…”

Section: The General Strategy Of Causality-based Feature Selectionmentioning

confidence: 99%

See 2 more Smart Citations

Causality-based Feature Selection: Methods and Evaluations

Yu,

Guo,

Liu

et al. 2019

Preprint

View full text Add to dashboard Cite

Feature selection is a crucial preprocessing step in data analytics and machine learning. Classical feature selection algorithms select features based on the correlations between predictive features and the class variable and do not attempt to capture causal relationships between them. It has been shown that the knowledge about the causal relationships between features and the class variable has potential benefits for building interpretable and robust prediction models, since causal relationships imply the underlying mechanism of a system. Consequently, causality-based feature selection has gradually attracted greater attentions and many algorithms have been proposed. In this paper, we present a comprehensive review of recent advances in causality-based feature selection. To facilitate the development of new algorithms in the research area and make it easy for the comparisons between new methods and existing ones, we develop the first open-source package, called CausalFS, which consists of most of the representative causality-based feature selection algorithms (available at https://github.com/kuiy/CausalFS). Using CausalFS, we conduct extensive experiments to compare the representative algorithms with both synthetic and real-world data sets. Finally, we discuss some challenging problems to be tackled in future causality-based feature selection research.

show abstract

Section: Overview Of Constraint-based Methodsmentioning

confidence: 99%

Section: Methods Of Mb Learning With Relaxed Assumptionsmentioning

confidence: 99%

Section: The General Strategy Of Causality-based Feature Selectionmentioning

confidence: 99%

See 1 more Smart Citation

Causality-based Feature Selection: Methods and Evaluations

Yu,

Guo,

Liu

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Tsamardinos et al (2003b) improved on this work by adding those variables to the candidate MB having the strongest dependencies with the target in advance. While employing the same statistical tests and heuristics, another algorithm that learns the direct neighbors and spouses separately has proven superior, and hence has been widely adopted in later constraint-based methods (see Aliferis et al (2003); Tsamardinos et al (2003a); Peña et al (2007); Fu and Desmarais (2008); Aliferis et al (2010a); de Morais and Aussem (2010); Liu and Liu (2016)). In particular, this improved strategy looks at the dependencies with the target at a distance of one and two neighbors separately.…”

Section: Related Workmentioning

confidence: 99%

Markov Blanket Discovery using Minimum Message Length

Li¹,

Korb

Allison³

2021

Preprint

View full text Add to dashboard Cite

Causal discovery automates the learning of causal Bayesian networks from data and has been of active interest from their beginning. With the sourcing of large data sets off the internet, interest in scaling up to very large data sets has grown. One approach to this is to parallelize search using Markov Blanket (MB) discovery as a first step, followed by a process of combining MBs in a global causal model. We develop and explore three new methods of MB discovery using Minimum Message Length (MML) and compare them empirically to the best existing methods, whether developed specifically as MB discovery or as feature selection. Our best MML method is consistently competitive and has some advantageous features.

show abstract

“…The experimental assembly and high-P experimental technique were generally identical to those reported in. 21 Our sample was synthesized at 1 GPa and 800 C with a heating time of 24 hours.…”

Section: Synthesismentioning

confidence: 99%