🧠 VAE Explorer
Interactive Variational Autoencoder - Learn how VAEs encode, sample, and generate data
🏗️
VAE Architecture
Input Data
X (Original)
→
Encoder
Neural Network
→
μ (mean)
Latent Mean
σ (variance)
Latent Variance
→
Latent Space
z = μ + σ * ε
→
Decoder
Neural Network
→
Reconstruction
X' (Generated)
🌌
Latent Space Visualization (2D)
Left: Before Training (Random) | Right: After Training (Structured)
🎛️ Latent Dimensions
📊 Loss Metrics
Reconstruction Loss
0.234
KL Divergence
0.156
Total Loss
0.390
Accuracy
92.5%
⚙️ Training Parameters
✨
Generated Samples from Latent Space
🔄
Input vs Reconstruction
📈
0 Comments