alessiozanga / causal-hub Goto Github PK

Add Graph trait with the following:

• Clone trait bound,
• Debug trait bound,
• Display trait bound,
• Vertex associated type,
• Edge associated type,
• Order getter,
• Vertices iterator,
• Has vertex method,
• Add vertex by label method,
• Delete vertex method,
• Size getter,
• Edges iterator,
• Has edge method,
• Add edge method,
• Delete edge method,
• Is adjacent to method.

Add the following:

• CorrelationMatrix struct,
• CovarianceMatrix struct,
• PrecisionMatrix struct.

Add DirectedDenseMatrixGraph with the following:

• BaseGraph trait implementation,
• DefaultGraph trait implementation,
• PartialOrdGraph trait implementation,
• DirectedGraph trait implementation.

Add the following:

• Add AdjacencyList format,
• Add Edgelist format,
• Add GEXF format,
• #145
• Add GraphML format,
• Add Pajek format.

Add the following:

• Independence trait with method is_independent(X, Y, Z) to check for conditional independence of X _||_ Y | Z,
• Implement Independence for UndirectedGraph in terms of u-separation,
• Implement Independence for DirectedGraph in terms of d-separation.

Contact Details

[email protected]

What happened?

The pull_request_template.md is not detected by GitHub.

Relevant log output

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

Implement special cases of OLS, such as with 1, 2 and 3 regressors, to improve performance in low dimensionality.

Add the following:

• StARS algorithm as described here.

What happened?

The default size of a complete undirected graph with n labels is set for default to (n * (n+1)) / 2, while it must be set to n * (n+1) / 2 - n. Indeed this quantity is equivalent to (n * (n -1)) / 2, but it prevents the code to overflow in case of n = 0, with n being of type usize.

Code of Conduct

• I agree to follow this project's Code of Conduct

Contact Details

No response

What feature do you propose?

Add the constraint-based PC-Stable causal discovery algorithm.

In particular, the issue will be split in:

• Add the skeleton function of PC-Stable to get an undirected graph,
• Add the orient function of PC-Stable to apply orientation rules.
• Enable PC-Stable to include prior knowledge
• Benchmarks

Relevant references

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

Add the following:

• dtolnay/rust-toolchain action,
• cargo sort -c, for Cargo.toml formatting,
• cargo +nightly rustdoc -- --show-coverage -Z unstable-options, for docs coverage report.

Add the following:

• PartialCorrelation functor,
• ConditionalIndependenceTest trait,
• FisherZ CIT functor,
• StudentT CIT functor.

Add Dataset trait with the following:

• From and Into variables names and ndarray matrix,
• From and Into given polars dataframe,
• Check if there are missing values and build a "missing mask",
• Getter on the names of the variables,
• Getter on the underlying data matrix.

Add DiscreteDataset struct with the following:

• Implement Dataset trait,
• Getter on the levels of a variable,
• Getter on the cardinality vector.

Add ContinuousDataset struct with the following:

• Implement Dataset trait.

Add the following:

• 1D MarginalCountMatrix,
• 2D ConditionalCountMatrix,
• 3D JointCountMatrix.

Add the following:

• Models:
  • #75
  • Add Gaussian Bayesian network model
• Parsing:
  • #80
• Sampling:
  • #124
  • Add Rejection sampling
  • Add LikelihoodWeighting sampling
• Fitting:
  • #84
  • #86
  • Add ExpectationMaximization estimator
• Approximate Inference:
  • Add approximate inference
• Exact Inference
  • #78
  • #107

Add Default trait with the following:

• Graph trait bound,
• Default trait bound,
• Null constructor,
• Empty constructor,
• Complete constructor.

Add the following:

• Add ROC metric,
• Add AUC metric,
• Add From trait for graphs.

Add subgraphs and clusters following the documentation.

Add the traversal algorithms submodule with the following:

• Breadth First Search (BFS),
• Depth First Search (DFS),
• Lexicographical BFS (LexBFS),
• Lexicographical DFS (LexDFS),
• Topological sort.

Add the HillClimbing algorithm with the following:

• with_graph, to provide a non-empty initial DAG,
• with_max_iter, to limit the number of maximum iterations,
• with_max_parents, to limit the number of maximum parents for each vertex.

Add UndirectedGraph trait with the following:

• Graph trait bound,
• NeighborsIter associated type,
• Neighbors iterator,
• Is neighbor of method,
• Degree of method.

Add the following:

• Add README with status badges and sections,

• Add CHANGELOG,

• Add CODE_OF_CONDUCT,

• Add CONTRIBUTING,

• Add issue and pull request templates,

• Add CITATION.cff.

Add UndirectedDenseMatrixGraph with the following:

• AdjacencyMatrix crate-wide type,
• BaseGraph trait implementation,
• DefaultGraph trait implementation,
• PartialOrdGraph trait implementation,
• UndirectedGraph trait implementation.

Add the following:

• Iterator over the same edges w.r.t. given graph,
• Iterator over the extra edges w.r.t. given graph,
• Iterator over the missing edges w.r.t. given graph,
• Iterator over the reversed edges w.r.t. given graph,
• Structural Hamming Distance (SHD).

Add DirectedGraph trait with the following:

• AncestorsIter associated type,
• Ancestors iterator,
• Is ancestor of method,
• ParentsIter associated type,
• Parents iterator,
• Is parent of method,
• ChildrenIter associated type,
• Children iterator,
• Is child of method,
• DescendantsIter associated type,
• Descendants iterator,
• Is descendant of method.

Add PartialOrd trait with the following:

• Graph trait bound,
• Eq trait bound,
• PartialOrd trait bound,
• Is subgraph method,
• Is supergraph method.

Add IndexBaseGraph trait with the following:

• Unsafe getters and setters by index from BaseGraph.
• Unsafe getters and setters by index for other traits, such as UndirectedGraph.

Add projection of a (discrete) Bayesian network onto another.

• Add cargo check,
• Add cargo build,
• Add cargo test,
• Add clippy check,
• Add fmt check,
• Add cov coverage,
• Add cargo publish.

• Change Hill-Climbing functor to use immutable reference to scoring criterion instead of requiring ownership.