2021/10/28 by Plabon Shaha, Shaha, Plabon, Talha Islam Zadid +5
Computer Science · Mathematics · #Advanced Causal Inference Techniques #Artificial Intelligence (cs.AI) #Bayesian Modeling and Causal Inference #FOS: Computer and information sciences #Imbalanced Data Classification Techniques #Machine Learning (cs.LG) #Machine Learning and Data Classification #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.2110.15075
openalex publication_date 2021/10/28 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Estimating causal effects from observational data informs us about which\nfactors are important in an autonomous system, and enables us to take better\ndecisions. This is important because it has applications in selecting a\ntreatment in medical systems or making better strategies in industries or\nmaking better policies for our government or even the society. Unavailability\nof complete data, coupled with high cardinality of data, makes this estimation\ntask computationally intractable. Recently, a regression-based weighted\nestimator has been introduced that is capable of producing solution using\nbounded samples of a given problem. However, as the data dimension increases,\nthe solution produced by the regression-based method degrades. Against this\nbackground, we introduce a neural network based estimator that improves the\nsolution quality in case of non-linear and finitude of samples. Finally, our\nempirical evaluation illustrates a significant improvement of solution quality,\nup to around 55 %, compared to the state-of-the-art estimators.\n