Function hill_climb

rl3v2/visualize_hill_climbing.py:10–32

(
    f, bounds, pop_size=1, sigma=0.3, alpha=0.3, iterations=100
)

# Evolution Strategies optimizer (simple version)
def hill_climb(
    f, bounds, pop_size=1, sigma=0.3, alpha=0.3, iterations=100
):
    dim = 2
    mu = np.random.uniform(bounds[0], bounds[1], size=dim)
    
    history = []
    best_f = f(mu)

    for gen in range(iterations):
        # Sample noise
        noise = np.random.randn(pop_size, dim)
        population = mu + sigma * noise
        fitness = np.array([f(x[0], x[1]) for x in population])

        history.append((population.copy(), mu.copy()))

        # Update point if it's better
        if fitness[0] > best_f:
            best_f = fitness[0]
            mu = population.flatten()
    
    return history

# Visualization function
def visualize_es(history, bounds, f, resolution=100):