Latent Space in AI

What I worked on

Tried to understand what latent space actually represents and how it connects to feature learning and the Alberta Plan ideas. Looked at where latent space exists in neural networks.

What I noticed

• Latent space is the internal learned representation of input data
• It typically exists between network layers
• The Alberta Plan focuses on learning which features matter most

”Aha” Moment

Latent space is the model’s internal world — its way of compressing meaning from raw data.

What still feels messy

What parts of a model contribute most to structure in latent space and how to measure “goodness” of representations.

Next step

Compare different architectures to see how latent spaces differ visually.