Filters








1,227 Hits in 7.1 sec

Learning Functional Causal Models with Generative Neural Networks [article]

Olivier Goudet, Diviyan Kalainathan, Philippe Caillou, Isabelle Guyon, David Lopez-Paz, Michèle Sebag
2017 arXiv   pre-print
Secondly, CGNN is applied to the problem of identifying v-structures and conditional independences.  ...  CGNN leverages the power of neural networks to learn a generative model of the joint distribution of the observed variables, by minimizing the Maximum Mean Discrepancy between generated and observed data  ...  We use the order-independent constraint-based version proposed by Colombo and Maathuis (2014) and the majority rules for the orientation of the edges.  ... 
arXiv:1709.05321v2 fatcat:5a5baqasizfxbejkbslf5z2z4y

Causal Generative Neural Networks [article]

Olivier Goudet, Diviyan Kalainathan, Philippe Caillou, Isabelle Guyon, David Lopez-Paz, Michèle Sebag
2018 arXiv   pre-print
CGNNs leverage conditional independencies and distributional asymmetries to discover bivariate and multivariate causal structures.  ...  We present Causal Generative Neural Networks (CGNNs) to learn functional causal models from observational data.  ...  Constraints based method PC with powerful HSIC conditional independence test is the second best performing method.  ... 
arXiv:1711.08936v2 fatcat:ehvaabpkd5d4vmi22s4igaon4y

Scalable Causal Structure Learning: New Opportunities in Biomedicine [article]

Pulakesh Upadhyaya, Kai Zhang, Can Li, Xiaoqian Jiang, Yejin Kim
2021 arXiv   pre-print
We review prominent traditional, score-based and machine-learning based schemes for causal structure discovery, study some of their performance over some benchmark datasets, and discuss some of the applications  ...  This paper gives a practical tutorial on popular causal structure learning models with examples of real-world data to help healthcare audiences understand and apply them.  ...  SAM (Structural Agnostic Modeling) Structurally agnostic model for causal discovery and penalized adversarial learning.  ... 
arXiv:2110.07785v1 fatcat:3dk2kfkvzjdqhazuenuhvg5f7e

On the Role of Sparsity and DAG Constraints for Learning Linear DAGs [article]

Ignavier Ng, AmirEmad Ghassami, Kun Zhang
2021 arXiv   pre-print
In this paper, we study the asymptotic role of the sparsity and DAG constraints for learning DAG models in the linear Gaussian and non-Gaussian cases, and investigate their usefulness in the finite sample  ...  Using gradient-based optimization and GPU acceleration, our procedure can easily handle thousands of nodes while retaining a high accuracy.  ...  Two major classes of structure learning methods are constraint-and score-based methods.  ... 
arXiv:2006.10201v3 fatcat:2ucie55ffbdqzjwkkpiht2i3ra

Physical System for Non Time Sequence Data [article]

Xiongren Chen
2020 arXiv   pre-print
We propose a novelty approach to connect machine learning to causal structure learning by jacobian matrix of neural network w.r.t. input variables.  ...  By functions fitting with Neural ODE, we can read out causal structure from functions.  ...  Related Work Traditionally, there are three main families of methods for causal structure learning, namely, constraint-based methods, score-based methods and structural causal function model-based methods  ... 
arXiv:2010.03206v1 fatcat:uuea7zxtpzfjhgbddlgsfppo6q

ParaLiNGAM: Parallel Causal Structure Learning for Linear non-Gaussian Acyclic Models [article]

Amirhossein Shahbazinia, Saber Salehkaleybar, Matin Hashemi
2021 arXiv   pre-print
In linear non-Gaussian acyclic models (LiNGAM), it can be shown that the true underlying causal structure can be identified uniquely from merely observational data.  ...  In this paper, we propose a parallel algorithm, called ParaLiNGAM, to learn casual structures based on DirectLiNGAM algorithm.  ...  Causal Structure Learning Algorithms for LiNGAM ICA-LiNGAM [12] was the first algorithm for the LiNGAM model, which applies an independent component analysis (ICA) algorithm to observed data and try  ... 
arXiv:2109.13993v1 fatcat:7rdxdlrtqvczvd6ombvzzstuse

D'ya like DAGs? A Survey on Structure Learning and Causal Discovery [article]

Matthew J. Vowels, Necati Cihan Camgoz, Richard Bowden
2021 arXiv   pre-print
Finally, we discuss the assumptive leap required to take us from structure to causality.  ...  Causal reasoning is a crucial part of science and human intelligence. In order to discover causal relationships from data, we need structure discovery methods.  ...  Constraint-Based and Score-Based Approaches Most constraint-based approaches test for conditional independencies in the empirical joint distribution in order to construct a graph that reflects these conditional  ... 
arXiv:2103.02582v2 fatcat:x45blijl5ze5xjyuqh6vlc26oq

D'ya Like DAGs? A Survey on Structure Learning and Causal Discovery

Matthew J. Vowels, Necati Cihan Camgoz, Richard Bowden
2022 ACM Computing Surveys  
Finally, we discuss the assumptive leap required to take us from structure to causality.  ...  Causal reasoning is a crucial part of science and human intelligence. In order to discover causal relationships from data, we need structure discovery methods.  ...  Constraint-Based and Score-Based Approaches Most constraint-based approaches test for conditional independencies in the empirical joint distribution in order to construct a graph that relects these conditional  ... 
doi:10.1145/3527154 fatcat:sroohzvx5reajkia5ythaiyyjm

Load-Balanced Parallel Constraint-Based Causal Structure Learning on Multi-Core Systems for High-Dimensional Data

Christopher Schmidt, Johannes Huegle, Philipp Bode, Matthias Uflacker
2019 Knowledge Discovery and Data Mining  
In the context of high-dimensional data state-of-the-art methods for constraint-based causal structure learning, such as the PC algorithm, are limited in their application through their worst case exponential  ...  In our work, we propose a parallel implementation that follows a dynamic task distribution in order to avoid situations of load imbalance and improve the execution time.  ...  Acknowledgments The authors would like to thank Hendrik Raetz, Frederic Schneider and Nils Thamm for helping with the implementation of the parallel adjacency search utilizing a centralized queue.  ... 
dblp:conf/kdd/SchmidtHBU19 fatcat:x3agmeij3ncbxpru3bdprqqhq4

Structural Agnostic Modeling: Adversarial Learning of Causal Graphs [article]

Diviyan Kalainathan, Olivier Goudet, Isabelle Guyon, David Lopez-Paz, Michèle Sebag
2022 arXiv   pre-print
A learning criterion combining distribution estimation, sparsity and acyclicity constraints is used to enforce the optimization of the graph structure and parameters through stochastic gradient descent  ...  A new causal discovery method, Structural Agnostic Modeling (SAM), is presented in this paper.  ...  Mikael Escobar-Bach for proofreading the paper. This work was granted access to the HPC resources of CCIPL (Nantes, France).  ... 
arXiv:1803.04929v5 fatcat:ynkjkpqj6vdvplnkzfbjgntt24

MissDAG: Causal Discovery in the Presence of Missing Data with Continuous Additive Noise Models [article]

Erdun Gao, Ignavier Ng, Mingming Gong, Li Shen, Wei Huang, Tongliang Liu, Kun Zhang, Howard Bondell
2022 arXiv   pre-print
In the M-step, MissDAG leverages the density transformation to model the noise distributions with simpler and specific formulations by virtue of the ANMs and uses a likelihood-based causal discovery algorithm  ...  We demonstrate the flexibility of MissDAG for incorporating various causal discovery algorithms and its efficacy through extensive simulations and real data experiments.  ...  C.2 Soft constraints GOLEM [35] employs likelihood-based objective with soft sparsity and DAG constraints for structure learning.  ... 
arXiv:2205.13869v1 fatcat:mngycou2kbfffiqzl5pqubfm4u

DiBS: Differentiable Bayesian Structure Learning [article]

Lars Lorch, Jonas Rothfuss, Bernhard Schölkopf, Andreas Krause
2021 arXiv   pre-print
Using DiBS, we devise an efficient, general purpose variational inference method for approximating distributions over structural models.  ...  Bayesian structure learning allows inferring Bayesian network structure from data while reasoning about the epistemic uncertainty -- a key element towards enabling active causal discovery and designing  ...  We thank Nicolo Ruggeri and Guillaume Wang for their valuable feedback.  ... 
arXiv:2105.11839v3 fatcat:qqhh6sljsfe3vlqknk5mkk2ahm

Learning Latent Causal Dynamics [article]

Weiran Yao, Guangyi Chen, Kun Zhang
2022 arXiv   pre-print
One critical challenge of time-series modeling is how to learn and quickly correct the model under unknown distribution shifts.  ...  The correction step is then formulated as learning the low-dimensional change factors with a few samples from the new environment, leveraging the identified causal structure.  ...  For this task, constraint-based methods (Entner & Hoyer, 2010) apply the conditional independence tests to recover the causal structures, while score-based methods (Murphy et al., 2002; Pamfil et al  ... 
arXiv:2202.04828v4 fatcat:6ejljwyd4nadbpntr6k4v7m3bi

Masked Gradient-Based Causal Structure Learning [article]

Ignavier Ng, Shengyu Zhu, Zhuangyan Fang, Haoyang Li, Zhitang Chen, Jun Wang
2022 arXiv   pre-print
We then utilize the reformulated SEM to develop a causal structure learning method that can be efficiently trained using gradient-based optimization, by leveraging a smooth characterization on acyclicity  ...  This paper studies the problem of learning causal structures from observational data.  ...  In this work, we develop a gradient-based optimization framework for structure learning, called Masked gradient-based Causal Structure Learning (MCSL), which (1) flexibly includes different model functions  ... 
arXiv:1910.08527v3 fatcat:u2u27nwcrvfkjo3jzrzers5gqe

A Bregman Method for Structure Learning on Sparse Directed Acyclic Graphs [article]

Manon Romain, Alexandre d'Aspremont
2020 arXiv   pre-print
We develop a Bregman proximal gradient method for structure learning on linear structural causal models.  ...  While the problem is non-convex, has high curvature and is in fact NP-hard, Bregman gradient methods allow us to neutralize at least part of the impact of curvature by measuring smoothness against a highly  ...  Related Work Structure Learning methods historically divide into constraint-based methods that test for conditional independence relations and score-based methods that optimize a variety of heuristics.  ... 
arXiv:2011.02764v1 fatcat:36qai7a6z5gjrkq6yy2nhik7c4
« Previous Showing results 1 — 15 out of 1,227 results