optimize_cmaes
Read-only
Optimize continuous parameters in non-convex, noisy, or gradient-free objectives using CMA-ES. Ideal for hyperparameter tuning, simulator calibration, and control policy optimization.
Instructions
[Premium] Continuous black-box optimization via CMA-ES (Covariance Matrix Adaptation Evolution Strategy). Use for tuning N continuous parameters when the objective is non-convex, noisy, or has no gradient — hyperparameter search, simulator calibration, control policy tuning. 10-100x fewer evaluations than grid search. For discrete combinatorial problems, use optimize_evolve. For LP/MIP problems with linear constraints, use solve_constraints. Stochastic init means re-runs with the same input may differ slightly. Requires ORACLAW_API_KEY.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| dimension | Yes | Number of parameters to optimize. | |
| objectiveWeights | Yes | Per-dimension weight in the linear default objective. Length must equal dimension. | |
| initialSigma | No | Initial step size (default: 0.5). | |
| maxIterations | No | Max generations (default: 1000, capped at 5000). | |
| initialMean | No | Optional starting point in parameter space. |
Output Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| bestSolution | Yes | Best parameter vector found. | |
| bestFitness | Yes | Objective value at bestSolution (caller's sign convention). | |
| iterations | Yes | Generations actually run. | |
| evaluations | No | Total objective evaluations. | |
| converged | Yes | Whether convergence criteria were met before maxIterations. | |
| executionTimeMs | No |