HOLOS — a fast, dependency-free Hyperdimensional Computing (HDC / VSA) engine.

HDC represents information as very high-dimensional vectors (~10,000-D) and computes with three cheap algebraic operations plus an associative memory:

• bind (⊗): associate two hypervectors (XOR) — invertible, result orthogonal to both.
• bundle (⊕): superpose several (majority) — result similar to all of them.
• permute (ρ): rotate — encode order / sequences.
• cleanup: nearest-neighbor search over an [ItemMemory] to decode a noisy result.

The flagship [Hypervector] uses the BSC (Binary Spatter Code) model, bit-packed into u64 words so operations become XOR + hardware popcount — very fast and cache-friendly. A bipolar [map::MapVector] model is also provided for tasks that favor linear bundling.

use holos_core::{Hypervector, Rng};
let mut rng = Rng::new(42);
let d = 10_000;
let role = Hypervector::random(d, &mut rng);
let value = Hypervector::random(d, &mut rng);
let bound = role.bind(&value);                 // associate role↔value
assert!(bound.bind(&role).similarity(&value) > 0.99); // unbind recovers value