Skip to content

Causal_Discovery_Algorithms.md

Latest commit

 

History

History
114 lines (73 loc) · 8.74 KB

Causal_Discovery_Algorithms.md

File metadata and controls

114 lines (73 loc) · 8.74 KB
graph LR
    Constraint_Based_Algorithms["Constraint-Based Algorithms"]
    FCM_Based_Algorithms["FCM-Based Algorithms"]
    Permutation_Based_Algorithms["Permutation-Based Algorithms"]
    Hidden_Causal_Algorithms["Hidden Causal Algorithms"]
    Score_Based_Algorithms["Score-Based Algorithms"]
    Graph_Representation_Layer["Graph Representation Layer"]
    Causal_Test_Utilities["Causal Test Utilities"]
    Constraint_Based_Algorithms -- "relies on" --> Causal_Test_Utilities
    Constraint_Based_Algorithms -- "depends on" --> Graph_Representation_Layer
    FCM_Based_Algorithms -- "depends on" --> Graph_Representation_Layer
    FCM_Based_Algorithms -- "may utilize" --> Causal_Test_Utilities
    Permutation_Based_Algorithms -- "depends on" --> Graph_Representation_Layer
    Permutation_Based_Algorithms -- "interacts with" --> Causal_Test_Utilities
    Hidden_Causal_Algorithms -- "interacts with" --> Causal_Test_Utilities
    Hidden_Causal_Algorithms -- "depends on" --> Graph_Representation_Layer
    Score_Based_Algorithms -- "interacts with" --> Causal_Test_Utilities
    Score_Based_Algorithms -- "depends on" --> Graph_Representation_Layer
    Graph_Representation_Layer -- "supports" --> Constraint_Based_Algorithms
    Graph_Representation_Layer -- "supports" --> FCM_Based_Algorithms
    Graph_Representation_Layer -- "supports" --> Permutation_Based_Algorithms
    Graph_Representation_Layer -- "supports" --> Hidden_Causal_Algorithms
    Graph_Representation_Layer -- "supports" --> Score_Based_Algorithms
    Causal_Test_Utilities -- "provides services to" --> Constraint_Based_Algorithms
    Causal_Test_Utilities -- "provides services to" --> Score_Based_Algorithms
    Causal_Test_Utilities -- "provides services to" --> Hidden_Causal_Algorithms
    Causal_Test_Utilities -- "provides services to" --> Permutation_Based_Algorithms
    Causal_Test_Utilities -- "may provide services to" --> FCM_Based_Algorithms
    click Graph_Representation_Layer href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main/causal-learn/Graph_Representation_Layer.md" "Details"
Loading

CodeBoardingDemoContact

Details

The causal-learn project is structured around various causal discovery algorithms, categorized by their underlying principles: Constraint-Based, FCM-Based, Permutation-Based, Hidden Causal, and Score-Based. These algorithms primarily interact with a central Graph Representation Layer for managing causal graph structures and rely on a Causal Test Utilities component for statistical tests and score calculations. This modular design allows for independent development and extension of different causal discovery methodologies while ensuring a consistent approach to graph manipulation and statistical inference.

Constraint-Based Algorithms

This component implements causal discovery methods that rely on conditional independence tests to construct and orient causal graphs. It includes algorithms like PC and FCI, which are fundamental to this paradigm. It also implicitly includes helper functions specific to these algorithms.

Related Classes/Methods:

FCM-Based Algorithms

This component groups algorithms that infer causal relationships based on Functional Causal Models (FCMs). This includes methods like Additive Noise Models (ANM), Linear Non-Gaussian Acyclic Models (LiNGAM), and Post-Nonlinear (PNL) models, which leverage specific assumptions about the functional relationship between cause and effect.

Related Classes/Methods:

Permutation-Based Algorithms

This component focuses on causal discovery by searching through permutations of variables to identify causal orderings. Algorithms like BOSS and GRaSP fall into this category, often optimizing a score function during the search process.

Related Classes/Methods:

Hidden Causal Algorithms

This component implements causal discovery methods specifically designed to address the challenge of hidden confounders. It focuses on learning latent representations and performing causal inference on these representations, exemplified by the GIN algorithm.

Related Classes/Methods:

Score-Based Algorithms

This component contains algorithms that discover causal graphs by optimizing a predefined score function. These methods search for the graph structure that best fits the observed data according to the chosen scoring criterion, including algorithms like CALM and Exact Search.

Related Classes/Methods:

Graph Representation Layer [Expand]

This component provides the fundamental data structures and operations for representing and manipulating causal graphs. It ensures a consistent and unified way for all causal discovery algorithms to interact with graph objects.

Related Classes/Methods:

Causal Test Utilities

This component centralizes various statistical tests and score calculation functions that are commonly used across different causal discovery algorithms. This includes conditional independence tests, which are crucial for constraint-based methods, and graph score evaluation functions for score-based methods.

Related Classes/Methods:

FAQ