READMEmexican.md

Hypervector - Mexican Dollar

This example demonstrates analogy inference in hyperdimensional computing:

"What is the dollar of Mexico?"

Using hypervector algebra, the system infers that the answer is Mexican peso (mpe).

Idea

We encode relationships like:

    USA → (name, capital, currency)
    Mexico → (name, capital, currency)

These are stored as bundled hypervectors:

    USA     = name ⊗ usa + capital ⊗ wdc + currency ⊗ usd
    Mexico  = name ⊗ mex + capital ⊗ cdmx + currency ⊗ mpe

We then compute a transformation:

    T = Mexico ⊗ USA⁻¹

and apply it to usd:

    T ⊗ usd ≈ mpe

Finally, we use cleanup (nearest neighbor search) to recover the closest known symbol.

What this shows

• Binding (⊗) encodes relationships
• Bundling (+) aggregates facts
• Inverse (⁻¹) enables analogy transfer
• The same computation works across multiple HDV types

Preliminaries

Install Rust.

Example Run

cargo run --example mexican

Binary
=========
Nearest HDV is: mpe


Bipolar
=========
Nearest HDV is: mpe


Modular
=========
Nearest HDV is: mpe


ComplexHDV
=========
Nearest HDV is: mpe


RealHDV
=========
Nearest HDV is: mpe

Notes

• All hypervectors are randomly generated but reproducible (fixed seed).
• Different representations (binary, bipolar, complex, etc.) yield the same result.
• Dimensionality affects robustness but not the underlying logic.

References

1. "What We Mean When We Say 'What’s the Dollar of Mexico?'" – Pentti Kanerva, 2010