Project Primer: Vector Transmissions

Origins

Visual explorations of high-dimensional vector spaces using seven dimensionality-reduction techniques — PCA, t-SNE, MDS, UMAP, SPLOM and others — applied to scikit-learn's Digits Dataset. Builds on Words Fail, Send Love with the idea that art is communication in high-dimensional space.

Structure

• Python-based (Matplotlib, Seaborn, UMAP, scikit-learn)
• Seven major visualisations, each demonstrating trade-offs in preserving local vs. global structure
• Images hosted on Vercel Blob, displayed with ShimmerImage
• Narrative explores the concept that emotions, context, and experience occupy high-dimensional space

Agent Guidance

• Part of the "vector trilogy": Words Fail → Vector Transmissions → Vector Atlas
• The philosophical thread: if art communicates meaning that language cannot fully capture, then vector spaces are a natural medium for that communication
• Python-based, unlike most of the site's TypeScript projects