Examples

JAX-HDC includes runnable examples in the examples/ directory.

Basic Operations

Core HDC operations: binding, bundling, permutation, and similarity.

python examples/basic_operations.py

from jax_hdc import MAP
import jax

model = MAP.create(dimensions=10000)
key = jax.random.PRNGKey(42)
x, y = model.random(key, (2, 10000))

bound = model.bind(x, y)
bundled = model.bundle(jnp.stack([x, y]), axis=0)
sim = model.similarity(x, y)

Classification

End-to-end classification with synthetic data: encode → train → evaluate.

pip install -e ".[examples]"   # optional: for matplotlib, scikit-learn
python examples/classification_simple.py

from jax_hdc import MAP, RandomEncoder, CentroidClassifier

model = MAP.create(dimensions=10000)
encoder = RandomEncoder.create(
    num_features=20, num_values=10, dimensions=10000,
    vsa_model=model, key=jax.random.PRNGKey(42)
)
classifier = CentroidClassifier.create(
    num_classes=5, dimensions=10000, vsa_model=model
)

# Encode, fit, predict
key = jax.random.PRNGKey(42)
data = jax.random.randint(key, (100, 20), 0, 10)
labels = jax.random.randint(key, (100,), 0, 5)
encoded = encoder.encode_batch(data)
classifier = classifier.fit(encoded, labels)
predictions = classifier.predict(encoded)
accuracy = classifier.score(encoded, labels)

Kanerva’s “Dollar of Mexico”

Structured knowledge representation and analogical reasoning.

python examples/kanerva_example.py

See the HDC Learning tutorial for a step-by-step walkthrough of the classification pipeline.