@article{PAR00020756,
  title = {{U}sing unlabeled data to discover bivariate causality with deep restricted {B}oltzmann machines},
  author = {{S}okolovska, {N}. and {P}ermiakova, {O}. and {F}orslund, {S}. {K}. and {Z}ucker, {J}ean-{D}aniel},
  editor = {},
  language = {{ENG}},
  abstract = {{A}n important question in microbiology is whether treatment causes changes in gut flora, and whether it also affects metabolism. {T}he reconstruction of causal relations purely from non-temporal observational data is challenging. {W}e address the problem of causal inference in a bivariate case, where the joint distribution of two variables is observed. {W}e consider, in particular, data on discrete domains. {T}he state-of-the-art causal inference methods for continuous data suffer from high computational complexity. {S}ome modern approaches are not suitable for categorical data, and others need to estimate and fix multiple hyper-parameters. {I}n this contribution, we introduce a novel method of causal inference which is based on the widely used assumption that if {X} causes {Y}, then {P}({X}) and {P}({Y} vertical bar {X}) are independent. {W}e propose to explore a semi-supervised approach where {P}({Y} vertical bar {X}) and {P}({X}) are estimated from labeled and unlabeled data respectively, whereas the marginal probability is estimated potentially from much more (cheap unlabeled) data than the conditional distribution. {W}e validate the proposed method on the standard cause-effect pairs. {W}e illustrate by experiments on several benchmarks of biological network reconstruction that the proposed approach is very competitive in terms of computational time and accuracy compared to the state-of-the-art methods. {F}inally, we apply the proposed method to an original medical task where we study whether drugs confound human metagenome.},
  keywords = {{C}ausal inference ; semi-supervised learning ; probabilistic models ; metagenomic data},
  booktitle = {},
  journal = {{IEEE}-{ACM} {T}ransactions on {C}omputational {B}iology and {B}ioinformatics},
  volume = {17},
  numero = {1},
  pages = {358--364},
  ISSN = {1545-5963},
  year = {2020},
  DOI = {10.1109/tcbb.2018.2879504},
  URL = {https://www.documentation.ird.fr/hor/{PAR}00020756},
}