ContinuousOptimizationBase<T>
Base Classes ยท AiDotNet.CausalDiscovery.ContinuousOptimization
Base class for continuous optimization causal discovery methods (NOTEARS, DAGMA, GOLEM).
Inheritance: CausalDiscoveryBase<T> โ ContinuousOptimizationBase<T>
Known derived types: CORLAlgorithm<T>, DAGMALinear<T>, DAGMANonlinear<T>, GOLEMAlgorithm<T>, MCSLAlgorithm<T>, NOTEARSLinear<T>, NOTEARSLowRank<T>, NOTEARSNonlinear<T>, NOTEARSSobolev<T>, NoCurlAlgorithm<T>
For Beginners
Traditional methods check all possible graph structures to find the best one, which is extremely slow for many variables. Continuous optimization methods instead use calculus-based optimization (like gradient descent) to smoothly search for the best graph, which is much faster.
How It Works
These methods formulate the DAG learning problem as a continuous optimization: minimize a loss function (e.g., least squares) subject to a smooth acyclicity constraint. The key innovation is replacing the combinatorial DAG constraint with a differentiable function.
Properties
| Property | Summary |
|---|---|
CausalDiscoveryCategory Category { get; } |