Understanding Latent Space
• 1 min read 1 min
What I worked on
Looked into different ways to visualize and interpret latent space representations using PCA, t-SNE, and UMAP. Also tried to plot them in 3D with Plotly for interactive exploration.
What I noticed
- PCA, t-SNE, and UMAP reduce dimensionality effectively
”Aha” Moment
na
What still feels messy
Still not sure how to interpret cluster separations. Do they reflect learned structure or noise.
Next step
Run t-SNE and UMAP on different checkpoints to compare latent evolution.