Skip to content
TheChatArchiveThe Chat Archive
Kimikimi-k2-0711-preview· Moonshot AIHugging Face LMArena Human Preference 140K

Write a Python program that fits data with a number of Gaussian peaks and displays this.

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Metallama-4-scout-17b-16e-instruct’s answer.
  1. User

    Write a Python program that fits data with a number of Gaussian peaks and displays this.

  2. Kimi Assistant

    Below is a compact, self-contained script that

    1. generates synthetic data consisting of several Gaussian peaks plus noise,
    2. fits the data with a user-chosen number of Gaussians using non-linear least squares,
    3. plots the raw data, the individual fitted peaks and the total model.

    You can also feed it your own x, y vectors – just replace the synthetic-data block.

    #!/usr/bin/env python3
    """
    fit_n_gaussians.py
    Fit an arbitrary number of Gaussian peaks to 1-D data and display the result.
    """
    
    import numpy as np
    import matplotlib.pyplot as plt
    from scipy.optimize import curve_fit
    
    # ------------------------------------------------------------------
    # 1.  Synthetic data (replace with your own x, y if desired)
    # ------------------------------------------------------------------
    def make_fake_data():
        x = np.linspace(0, 100, 500)
        true_peaks = [
            (30, 1.0, 5),   # (center, amplitude, sigma)
            (50, 0.7, 7),
            (75, 0.9, 4)
        ]
        y = np.zeros_like(x)
        for c, a, s in true_peaks:
            y += a * np.exp(-0.5 * ((x - c) / s) ** 2)
        rng = np.random.default_rng(42)
        y += 0.05 * rng.normal(size=y.size)   # noise
        return x, y
    
    x, y = make_fake_data()
    
    # ------------------------------------------------------------------
    # 2.  Model: sum of n Gaussians
    # ------------------------------------------------------------------
    def gaussian(x, c, a, s):
        """Single Gaussian: center, amplitude, sigma."""
        return a * np.exp(-0.5 * ((x - c) / s) ** 2)
    
    def multi_gaussian(x, *params):
        """
        Sum of n Gaussians.
        params = [c1, a1, s1, c2, a2, s2, ...]
        """
        n = len(params) // 3
        out = 0.0
        for i in range(n):
            out += gaussian(x, *params[3*i:3*i+3])
        return out
    
    # ------------------------------------------------------------------
    # 3.  Fit
    # ------------------------------------------------------------------
    n_peaks = 3                       # <-- change this
    initial = []
    for k in range(n_peaks):
        # crude initial guesses: evenly spaced centers, equal amps, equal widths
        initial += [x.min() + (x.max()-x.min())*(k+0.5)/n_peaks,
                    y.max()/n_peaks,
                    (x.max()-x.min())/n_peaks/2]
    
    popt, pcov = curve_fit(multi_gaussian, x, y, p0=initial)
    
    # ------------------------------------------------------------------
    # 4.  Plot
    # ------------------------------------------------------------------
    plt.figure(figsize=(8, 5))
    plt.scatter(x, y, s=8, color='k', label='data')
    
    # total model
    x_fine = np.linspace(x.min(), x.max(), 1000)
    plt.plot(x_fine, multi_gaussian(x_fine, *popt), 'r-', lw=2, label='total fit')
    
    # individual peaks
    for i in range(n_peaks):
        c, a, s = popt[3*i:3*i+3]
        plt.plot(x_fine, gaussian(x_fine, c, a, s), '--', label=f'peak {i+1}')
    
    plt.xlabel('x')
    plt.ylabel('y')
    plt.legend()
    plt.title(f'Fit with {n_peaks} Gaussian peaks')
    plt.tight_layout()
    plt.show()
    

    How to use with real data

    Replace the synthetic-data block with something like:

    x = np.loadtxt('spectrum.txt')[:, 0]
    y = np.loadtxt('spectrum.txt')[:, 1]
    

    and adjust n_peaks to the number of peaks you expect.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.