ccevolve/tx_workspace/ac1_alphaevolve/solution_v5.py

import jax
import jax.numpy as jnp
import numpy as np
from scipy.optimize import minimize as scipy_minimize
import optax


def compute_c1_numpy(f_values, n_points):
    dx = 0.5 / n_points
    f_nn = np.maximum(f_values, 0.0)
    autoconv = np.convolve(f_nn, f_nn, mode='full') * dx
    integral_sq = (np.sum(f_nn) * dx) ** 2
    if integral_sq < 1e-12:
        return 1e10
    return np.max(autoconv) / integral_sq


def optimize_cascade(verbose=True):
    """Optimize with cascading resolution: start coarse, upsample, refine."""

    best_params_np = None

    for stage, (N, adam_steps, lr_peak) in enumerate([
        (1000, 60000, 0.01),
        (2000, 60000, 0.005),
        (4000, 80000, 0.003),
    ]):
        dx = 0.5 / N
        if verbose:
            print(f"\n=== Stage {stage}: N={N}, steps={adam_steps} ===")

        @jax.jit
        def objective_smooth(params, temp):
            f = jax.nn.softplus(params)  # smooth non-negativity, allows zeros
            padded = jnp.zeros(2 * N)
            padded = padded.at[:N].set(f)
            fft_f = jnp.fft.rfft(padded)
            conv = jnp.fft.irfft(fft_f * fft_f, n=2 * N) * dx
            integral_sq = (jnp.sum(f) * dx) ** 2
            smooth_max = jax.nn.logsumexp(temp * conv) / temp
            return smooth_max / integral_sq

        @jax.jit
        def objective_hard(params):
            f = jax.nn.softplus(params)
            padded = jnp.zeros(2 * N)
            padded = padded.at[:N].set(f)
            fft_f = jnp.fft.rfft(padded)
            conv = jnp.fft.irfft(fft_f * fft_f, n=2 * N) * dx
            integral_sq = (jnp.sum(f) * dx) ** 2
            return jnp.max(conv) / integral_sq

        grad_smooth = jax.jit(jax.grad(objective_smooth))

        # Initialize
        if best_params_np is None:
            np.random.seed(42)
            init_f = np.ones(N) * 0.5
            init_f += 0.02 * np.random.randn(N)
            # Inverse softplus
            params = jnp.array(np.log(np.expm1(np.maximum(init_f, 1e-4))))
        else:
            # Upsample from previous stage
            old_f = np.log1p(np.exp(best_params_np))  # softplus
            new_f = np.interp(np.linspace(0, 1, N), np.linspace(0, 1, len(old_f)), old_f)
            params = jnp.array(np.log(np.expm1(np.maximum(new_f, 1e-4))))

        # Adam optimization
        lr_schedule = optax.warmup_cosine_decay_schedule(
            init_value=0.0, peak_value=lr_peak, warmup_steps=2000,
            decay_steps=adam_steps - 2000, end_value=lr_peak * 1e-4,
        )
        optimizer = optax.adam(learning_rate=lr_schedule)
        opt_state = optimizer.init(params)

        best_c1 = float('inf')
        best_params = params

        for step in range(adam_steps):
            # Fixed high temperature
            temp = 200.0

            loss, grads = jax.value_and_grad(objective_smooth)(params, temp)
            updates, opt_state = optimizer.update(grads, opt_state, params)
            params = optax.apply_updates(params, updates)

            if step % 10000 == 0 or step == adam_steps - 1:
                hard_c1 = float(objective_hard(params))
                if verbose:
                    print(f"  Step {step:6d} | C1={hard_c1:.8f}")
                if hard_c1 < best_c1:
                    best_c1 = hard_c1
                    best_params = params

        # L-BFGS polishing
        if verbose:
            print(f"  L-BFGS polishing from C1={best_c1:.8f}")

        params_np = np.array(best_params, dtype=np.float64)
        for temp in [500.0, 2000.0, 10000.0]:
            def scipy_obj(p):
                p_jax = jnp.array(p)
                val = float(objective_smooth(p_jax, temp))
                g = np.array(grad_smooth(p_jax, temp), dtype=np.float64)
                return val, g

            result = scipy_minimize(
                scipy_obj, params_np, method='L-BFGS-B', jac=True,
                options={'maxiter': 5000, 'ftol': 1e-15, 'gtol': 1e-12},
            )
            params_np = result.x

            f_opt = np.log1p(np.exp(params_np))
            c1 = compute_c1_numpy(f_opt, N)
            if verbose:
                print(f"    temp={temp:.0f}: C1={c1:.10f}")
            if c1 < best_c1:
                best_c1 = c1
                best_params = jnp.array(params_np)

        best_params_np = np.array(best_params)
        if verbose:
            print(f"  Stage {stage} best: C1={best_c1:.10f}")

    # Final result
    f_final = np.log1p(np.exp(best_params_np))
    c1_final = compute_c1_numpy(f_final, N)
    return f_final, c1_final, N


def run():
    f, c1, N = optimize_cascade(verbose=True)
    print(f"\nFinal C1: {c1:.10f}")
    return f, c1, c1, N