CMA-ES · WebAssembly

Benchmark Playground

Pick a benchmark, set λ/σ, and watch CMA‑ES morph its Gaussian each generation. Log-loss, live scatter, covariance ellipse, scrubable timeline.

Bundle

Sequential (pkg)

Algorithm

CMA‑ES full covariance

Per-generation Gaussian, mean, covariance, σ adaptation.

Status

Idle

Benchmark

Lambda (pop)

Sigma

Max iters

Seed

Dimension (fixed)

2D surface view

Projection

Enable bounds

Noise model

Samples/pt

Max samples

Adaptive

Runs CMA‑ES (WASM) and renders the full per-generation timeline. Use the scrub bar to navigate.

WASM uses the Rust CMA-ES with SIMD (if enabled); chart drives the visible step loop on a 2D surface.

Bounds/noise settings apply to the optimizer; visualization shows the 2D slice.

Best f(t) –

3D surface + candidates Iter 0

Custom objective

Function signature: f(x: number[]): number. Keep it deterministic for fair races.