An interactive browser-based laboratory for exploring Hyperdimensional Computing (HDC) and Vector Symbolic Architectures (VSA) — a brain-inspired computing paradigm using ultra-high-dimensional random vectors (10,000-D) that bridges symbolic AI and neural computation.

1. HD Vector Playground — Generate random hypervectors, visualize near-orthogonality, explore binding/bundling/permutation operations, dimensionality sweep
2. Language Recognition — Encode character n-grams as hypervectors, single-pass language classification (English, Spanish, French, German, Italian, Portuguese)
3. Sparse Distributed Memory — Kanerva's SDM (NASA Ames, 1988): store/retrieve patterns with noise tolerance, capacity testing
4. Graph Encoding & Analogy — Role-filler bindings for knowledge graphs, analogy reasoning ("capital of France?")
5. Resonator Networks — Latest HDC breakthrough (Frady/Olshausen): iterative factorization of composed hypervectors
6. VSA vs Neural Network Arena — Head-to-head classification comparison with few-shot learning curves

• Binding (element-wise multiply): creates a vector dissimilar to both operands (encodes association)
• Bundling (element-wise add + threshold): creates a vector similar to all operands (encodes set membership)
• Permutation (cyclic shift): creates a dissimilar vector (encodes sequence/order)
• Near-orthogonality: random 10,000-D vectors are almost perpendicular — the "blessing of dimensionality"

• Single-pass learning — no backpropagation needed
• Energy efficient — simple operations (multiply, add, compare)
• Incremental — add new data by bundling, no retraining
• Interpretable — similarity-based reasoning
• Robust — tolerant to noise and hardware errors
• Active research at Intel Labs, IBM Research, ETH Zurich, UC Berkeley

• Single HTML file, zero dependencies
• Canvas 2D visualizations
• Pure JavaScript HDC engine
• Dark theme, responsive

• Kanerva, P. (1988). Sparse Distributed Memory. MIT Press.
• Plate, T. (2003). Holographic Reduced Representations. CSLI.
• Gayler, R. (2003). Vector Symbolic Architectures answer Jackendoff's challenges.
• Rahimi & Recht (2009). Weighted sums of random kitchen sinks.
• Joshi et al. (2016). Language recognition using random indexing.
• Frady et al. (2020). Resonator Networks for factoring high-dimensional representations.

MIT