CausalDiscoveryResult<T>
Models & Types · AiDotNet.CausalDiscovery
Contains the results of a causal discovery analysis, including the learned graph and convergence metrics.
For Beginners
After running causal discovery, this object tells you:
- The causal graph itself (which variables cause which others)
- Which algorithm was used
- How well the algorithm converged (did it find a good solution?)
- How sparse the graph is (fewer edges = more interpretable)
How It Works
This result object is populated when ConfigureCausalDiscovery() is used on the
AiModelBuilder. It contains the discovered causal graph along with algorithm-specific
metrics about the optimization process.
Constructors
| Constructor | Summary |
|---|---|
CausalDiscoveryResult(CausalGraph<T> graph, CausalDiscoveryAlgorithmType algorithmUsed, CausalDiscoveryCategory category, int iterations, double finalLoss, double acyclicityConstraint, bool converged, TimeSpan elapsedTime, Dictionary<string, double> additionalMetrics) |
Initializes a new CausalDiscoveryResult with full convergence metrics. |
Properties
| Property | Summary |
|---|---|
CausalDiscoveryAlgorithmType AlgorithmUsed { get; } |
Gets the algorithm that was used for discovery. |
CausalDiscoveryCategory Category { get; } |
Gets the category of the algorithm used. |
CausalGraph<T> Graph { get; } |
Gets the discovered causal graph (DAG). |
Dictionary<string, double> AdditionalMetrics { get; } |
Gets optional additional metrics specific to the algorithm used. |
TimeSpan ElapsedTime { get; } |
Gets the wall-clock time the algorithm took to run. |
bool Converged { get; } |
Gets whether the algorithm converged successfully. |
double AcyclicityConstraint { get; } |
Gets the final acyclicity constraint value h(W). |
double FinalLoss { get; } |
Gets the final loss/objective value at convergence. |
double GraphDensity { get; } |
Gets the graph density (fraction of possible edges present). |
int EdgeCount { get; } |
Gets the number of edges in the discovered graph. |
int Iterations { get; } |
Gets the number of iterations the algorithm performed. |