Causal Discovery

87 public types in this namespace, organized by kind.

Models & Types (76)

Type Summary
ANMAlgorithm<T> ANM (Additive Noise Model) — pairwise causal discovery via independence of residuals.
AVICIAlgorithm<T> AVICI — Amortized Variational Inference for Causal Discovery.
AmortizedCDAlgorithm<T> Amortized Causal Discovery — meta-learning approach to causal structure learning.
BCDNetsAlgorithm<T> BCD-Nets — Bayesian Causal Discovery Networks.
BOSSAlgorithm<T> BOSS (Best Order Score Search) — efficient permutation-based structure learning.
BayesDAGAlgorithm<T> BayesDAG — Bayesian DAG learning with gradient-based posterior inference.
CAMAlgorithm<T> CAM (Causal Additive Model) — order-based causal discovery with additive nonparametric regression.
CAMUVAlgorithm<T> CAM-UV — Causal Additive Model with Unobserved Variables.
CASTLEAlgorithm<T> CASTLE — Causal Structure Learning via neural networks with shared masked architecture.
CCDrAlgorithm<T> CCDr (Concave penalized Coordinate Descent with reparameterization) for DAG learning.
CCMAlgorithm<T> CCM — Convergent Cross-Mapping for detecting causation in nonlinear dynamical systems.
CDNODAlgorithm<T> CD-NOD — Constraint-based Discovery from Non-stationary / heterogeneous Data.
CGNNAlgorithm<T> CGNN — Causal Generative Neural Networks.
CORLAlgorithm<T> CORL — Causal Ordering via Reinforcement Learning.
CPCAlgorithm<T> CPC (Conservative PC) — PC variant that avoids erroneous v-structure orientation.
CausalDiscoveryResult<T> Contains the results of a causal discovery analysis, including the learned graph and convergence metrics.
CausalDiscoverySelector<T> Feature selector that uses any causal discovery algorithm to select features based on causal relationships.
CausalGraph<T> Represents a causal Directed Acyclic Graph (DAG) discovered from observational data.
CausalVAEAlgorithm<T> CausalVAE — Causal Variational Autoencoder.
DAGGNNAlgorithm<T> DAG-GNN — DAG Structure Learning with Graph Neural Networks.
DAGMALinear<T> DAGMA Linear — DAG learning via M-matrices and a log-determinant acyclicity characterization.
DAGMANonlinear<T> DAGMA Nonlinear — DAG learning via M-matrices and log-determinant with MLP structural equations.
DECIAlgorithm<T> DECI — Deep End-to-end Causal Inference.
DYNOTEARSAlgorithm<T> DYNOTEARS — Dynamic NOTEARS for time series structure learning.
DiBSAlgorithm<T> DiBS — Differentiable Bayesian Structure Learning.
DirectLiNGAMAlgorithm<T> DirectLiNGAM — direct method for LiNGAM without ICA.
ExactSearchAlgorithm<T> Exact Search (Dynamic Programming) — optimal DAG structure learning.
FCIAlgorithm<T> FCI (Fast Causal Inference) — constraint-based discovery with latent confounders.
FGESAlgorithm<T> FGES (Fast Greedy Equivalence Search) — greedy DAG search with BIC score caching.
FastIAMBAlgorithm<T> Fast-IAMB — faster variant of IAMB using speculative forward selection.
GAEAlgorithm<T> GAE — Graph Autoencoder for causal discovery.
GESAlgorithm<T> GES (Greedy Equivalence Search) — score-based causal discovery over equivalence classes.
GFCIAlgorithm<T> GFCI — Greedy FCI, a hybrid of GES and FCI.
GOBNILPAlgorithm<T> GOBNILP — Globally Optimal Bayesian Network learning using Integer Linear Programming.
GOLEMAlgorithm<T> GOLEM — Gradient-based Optimization with Likelihood for structure learning of linear DAGs.
GRaSPAlgorithm<T> GRaSP (Greedy Relaxation of Sparsest Permutation) — permutation-based causal discovery.
GraNDAGAlgorithm<T> GraN-DAG — Gradient-based Neural DAG Learning.
GrangerCausalityAlgorithm<T> Granger Causality — time series causal discovery via predictive improvement.
H2PCAlgorithm<T> H2PC — Hybrid HPC (Hybrid Parents and Children) algorithm.
HillClimbingAlgorithm<T> Hill Climbing — greedy score-based DAG structure learning.
IAMBAlgorithm<T> IAMB (Incremental Association Markov Blanket) — efficient Markov blanket discovery.
ICALiNGAMAlgorithm<T> ICA-LiNGAM — Linear Non-Gaussian Acyclic Model using Independent Component Analysis.
IGCIAlgorithm<T> IGCI (Information-Geometric Causal Inference) — bivariate causal discovery via entropy.
InterventionalDistribution<T> Represents the interventional distribution P(Y | do(X = x)) from Pearl’s do-calculus.
IterativeMCMCAlgorithm<T> Iterative MCMC — iteratively refined MCMC for Bayesian network structure learning.
K2Algorithm<T> K2 Algorithm — score-based learning with known variable ordering.
KraskovMIAlgorithm<T> Kraskov MI — Mutual Information estimation using k-nearest neighbors (KSG estimator).
LPCMCIAlgorithm<T> LPCMCI — Latent PCMCI for time series with hidden confounders.
MCSLAlgorithm<T> MCSL — Masked Gradient-Based Causal Structure Learning.
MMHCAlgorithm<T> MMHC — Max-Min Hill-Climbing, a hybrid constraint-based + score-based algorithm.
MMPCAlgorithm<T> MMPC (Max-Min Parents and Children) — identifies the parents and children of each variable.
MarkovBlanketAlgorithm<T> Markov Blanket (Grow-Shrink) Algorithm — discovers the Markov blanket of each variable.
NOTEARSLinear<T> NOTEARS Linear — continuous optimization for DAG structure learning with linear relationships.
NOTEARSLowRank<T> NOTEARS Low-Rank — DAG learning with low-rank parameterization for scalability.
NOTEARSNonlinear<T> NOTEARS Nonlinear — continuous optimization for DAG structure learning with nonlinear (MLP) relationships.
NOTEARSSobolev<T> NOTEARS with Sobolev regularization — DAG learning with smoothness constraints.
NTSNOTEARSAlgorithm<T> NTS-NOTEARS — Nonstationary Time Series NOTEARS.
NeuralGrangerAlgorithm<T> Neural Granger Causality — deep learning extension of Granger causality.
NoCurlAlgorithm<T> NoCurl — DAG learning via curl-free constraints on the graph structure.
OCSEAlgorithm<T> oCSE — optimal Causation Entropy for causal network inference.
OrderMCMCAlgorithm<T> Order MCMC — MCMC sampling over variable orderings for Bayesian structure learning.
PCAlgorithm<T> PC Algorithm — constraint-based causal discovery using conditional independence tests.
PCMCIAlgorithm<T> PCMCI — PC algorithm for Momentary Conditional Independence in time series.
PCMCIPlusAlgorithm<T> PCMCI+ — extension of PCMCI that also discovers contemporaneous causal links.
PCNOTEARSAlgorithm<T> PC-NOTEARS — Hybrid of PC skeleton discovery with NOTEARS continuous optimization.
PNLAlgorithm<T> PNL (Post-Nonlinear Causal Model) — Y = g(f(X) + N).
PartitionMCMCAlgorithm<T> Partition MCMC — MCMC sampling over DAG partitions for structure learning.
RCDAlgorithm<T> RCD (Repetitive Causal Discovery) — LiNGAM extension for latent confounders.
RFCIAlgorithm<T> RFCI (Really Fast Causal Inference) — scalable FCI for large datasets.
RSMAX2Algorithm<T> RSMAX2 — Restricted Maximization, a hybrid constraint-based + score-based algorithm.
TCDFAlgorithm<T> TCDF — Temporal Causal Discovery Framework.
TSFCIAlgorithm<T> tsFCI — time series Fast Causal Inference.
TabuSearchAlgorithm<T> Tabu Search — score-based DAG learning with memory to escape local optima.
TiMINoAlgorithm<T> TiMINo — Time series Models with Independent Noise (TiMINo causality, linear model class).
TransferEntropyAlgorithm<T> Transfer Entropy — information-theoretic measure of directed information flow.
VARLiNGAMAlgorithm<T> VAR-LiNGAM — Vector Autoregressive LiNGAM for time series causal discovery.

Base Classes (10)

Type Summary
BayesianCausalBase<T> Base class for Bayesian causal discovery algorithms.
CausalDiscoveryBase<T> Abstract base class for causal discovery algorithms with shared statistical utilities.
ConstraintBasedBase<T> Base class for constraint-based causal discovery algorithms (PC, FCI, MMPC, etc.).
ContinuousOptimizationBase<T> Base class for continuous optimization causal discovery methods (NOTEARS, DAGMA, GOLEM).
DeepCausalBase<T> Base class for deep learning-based causal discovery algorithms.
FunctionalBase<T> Base class for functional/ICA-based causal discovery algorithms (LiNGAM, ANM, etc.).
HybridBase<T> Base class for hybrid causal discovery algorithms that combine constraint-based and score-based methods.
InfoTheoreticBase<T> Base class for information-theoretic causal discovery algorithms.
ScoreBasedBase<T> Base class for score-based causal discovery algorithms (GES, FGES, Hill Climbing, Tabu, etc.).
TimeSeriesCausalBase<T> Base class for time series causal discovery algorithms (Granger, PCMCI, DYNOTEARS, etc.).

Interfaces (1)

Type Summary
ICausalDiscoveryAlgorithm<T> Interface for causal structure learning algorithms that discover Directed Acyclic Graphs (DAGs) from data.