Neural Network Weights

import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np
from matplotlib.colors import LinearSegmentedColormap

np.random.seed(42)

# Simulated neural network layer weights
input_neurons = [f'In_{i}' for i in range(16)]
hidden_neurons = [f'H_{i}' for i in range(12)]

# Xavier-like initialization with some structure
weights = np.random.randn(16, 12) * np.sqrt(2.0 / 16)
# Add learned patterns
weights[:4, :3] += 0.5
weights[4:8, 3:6] -= 0.5
weights[8:12, 6:9] += 0.3
weights[12:, 9:] -= 0.3

df = pd.DataFrame(weights, index=input_neurons, columns=hidden_neurons)

# NEON diverging colormap
neon_weights = LinearSegmentedColormap.from_list('neon_w', ['#27D3F5', '#0a0a0f', '#F5276C'])

g = sns.clustermap(df, cmap=neon_weights, center=0, 
                   method='ward', metric='euclidean',
                   linewidths=0.2, linecolor='#1a1a2e',
                   figsize=(8, 7), dendrogram_ratio=(0.12, 0.12),
                   cbar_pos=(0.01, 0.08, 0.008, 0.12),
                   tree_kws={'linewidths': 1.5, 'colors': '#27D3F5'})

g.fig.patch.set_facecolor('#0a0a0f')
g.ax_heatmap.set_facecolor('#0a0a0f')
g.ax_heatmap.tick_params(colors='white', labelsize=8)
g.ax_heatmap.set_xlabel('Hidden Neurons', color='white', fontsize=10)
g.ax_heatmap.set_ylabel('Input Neurons', color='white', fontsize=10)

for ax in [g.ax_row_dendrogram, g.ax_col_dendrogram]:
    ax.set_facecolor('#0a0a0f')

g.fig.suptitle('Neural Network Weight Matrix', color='white', fontsize=14, fontweight='bold', y=1.02)


plt.show()