Neural Network Loss Landscape

import matplotlib.pyplot as plt
import numpy as np
from matplotlib.colors import LinearSegmentedColormap

# Create complex optimization landscape
x = np.linspace(-3, 3, 250)
y = np.linspace(-3, 3, 250)
X, Y = np.meshgrid(x, y)

# Rastrigin-like function
A = 10
Z = A * 2 + (X**2 - A * np.cos(2 * np.pi * X)) + (Y**2 - A * np.cos(2 * np.pi * Y))
Z = np.log1p(Z)

# ML/AI dark theme
plt.style.use('dark_background')
fig, ax = plt.subplots(figsize=(10, 8))
fig.patch.set_facecolor('#0d1117')
ax.set_facecolor('#0d1117')

# Custom colormap - cool blues
colors = ['#0d1117', '#0a2a4a', '#0a4a7a', '#0a6aaa', '#2a8aca', 
          '#4aaadd', '#6acaee', '#8aeaff', '#aaffff', '#ffffff']
cmap = LinearSegmentedColormap.from_list('ml_blue', colors, N=256)

# Filled contour
cf = ax.contourf(X, Y, Z, levels=30, cmap=cmap)

# Subtle contour lines
cs = ax.contour(X, Y, Z, levels=15, colors='#8aeaff', linewidths=0.3, alpha=0.3)

# Mark global minimum with glow
ax.scatter([0], [0], color='#ff6b6b', s=200, marker='*', zorder=6, 
           edgecolors='white', linewidths=1.5)
ax.scatter([0], [0], color='#ff6b6b', s=400, marker='*', zorder=5, alpha=0.3)

# Gradient descent path
np.random.seed(42)
path_x, path_y = [2.5], [2.5]
for _ in range(35):
    dx = -0.12 * path_x[-1] + 0.015 * np.random.randn()
    dy = -0.12 * path_y[-1] + 0.015 * np.random.randn()
    path_x.append(path_x[-1] + dx)
    path_y.append(path_y[-1] + dy)

ax.plot(path_x, path_y, '-', color='#ffa502', linewidth=2, alpha=0.9, zorder=4)
ax.scatter(path_x[::5], path_y[::5], color='#ffa502', s=30, zorder=5, edgecolors='white', linewidths=0.5)

# Styled colorbar
cbar = plt.colorbar(cf, ax=ax, pad=0.02, shrink=0.85)
cbar.ax.set_facecolor('#0d1117')
cbar.set_label('log(Loss + 1)', fontsize=11, color='#c9d1d9', labelpad=10)
cbar.ax.yaxis.set_tick_params(color='#c9d1d9')
cbar.outline.set_edgecolor('#30363d')
plt.setp(plt.getp(cbar.ax.axes, 'yticklabels'), color='#c9d1d9', fontsize=9)

# Labels
ax.set_xlabel('Weight 1', fontsize=13, color='#c9d1d9', fontweight='600', labelpad=10)
ax.set_ylabel('Weight 2', fontsize=13, color='#c9d1d9', fontweight='600', labelpad=10)
ax.set_title('Neural Network Loss Landscape', fontsize=16, color='white', 
             fontweight='bold', pad=20)

# Style axes
ax.tick_params(colors='#8b949e', labelsize=10, length=0)
for spine in ax.spines.values():
    spine.set_visible(False)

ax.set_aspect('equal')
plt.tight_layout()
plt.show()