Monte Carlo Integration & Variance Reduction | Mathematical Demo by Markus MeighanMonte Carlo Integration & Variance Reduction | Mathematical Demo by Markus Meighan

Monte Carlo Integration & Variance Reduction | Mathematical Demo

Portfolio project
Markus Meighan
Independent synthetic demonstration — not commissioned client work.

The problem

Estimate the integral of exp(-sum(x_j)) over the 10-dimensional unit cube. Its exact value, (1-exp(-1))^10, provides an independent reference.

Mathematical approach

monte-carlo-latex.pdf
Compare plain Monte Carlo, antithetic pairs X and 1-X, and polynomial control variates based on the centered coordinate sum and its square. Coefficients use an independent 1,000-point pilot, included in the evaluation budget.

Results and validation

300 independent replications at five budgets (2,000–32,000 evaluations). At the largest budget, controls had 11.63 times lower empirical variance and 3.38 times lower RMSE than plain Monte Carlo. Approximate 95% interval coverage was 94.3%. Checks cover quadrature, seeded reproducibility and analytic variance formulas.

Limitations

This deliberately tractable, separable benchmark depends on the coordinate sum. Improvement is specific to this example, not a universal speedup. Equal evaluation budgets do not imply equal runtime.

Deliverables

Python/NumPy code, 4,500 replicate results, convergence figure and a three-page mathematical report. Methods reference: Art B. Owen, Monte Carlo theory, methods and examples, Chapter 8.

0 likes · 3 views · Nov 10, 2026

Synthetic Monte Carlo benchmark: exact solution, variance reduction, convergence and confidence-interval checks in Python.

Tags