Skip to main content
Featured image for post: Understanding Latent Space

Understanding Latent Space

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.