- Implemented the forward and backward passes manually in NumPy.
- Trained on Fashion MNIST and generated new images by sampling the latent space.
- Visualized latent-space structure to study disentanglement and KL-divergence behavior.
- Wrote a custom training loop to measure and optimize throughput.
10 / MACHINE LEARNING
VAE from scratch
A variational autoencoder with backpropagation written by hand — no PyTorch, no auto-grad library.
Code on GitHub ↗



