Tutorials

A read-in-order tour of the library. Each tutorial is a runnable .py file with copy-pasteable sections — no notebooks required, but every file opens cleanly in Google Colab too.

Available now

01_quickstart.py

Installation to first prediction in about ninety seconds: a first Gaussian hypervector, an iris classifier, temperature scaling and conformal prediction, and anomaly detection in five lines.

02_anomaly_detection.py

Calibrated one-shot anomaly detection from first principles. Shows the coverage guarantee empirically over 200 splits, the drift of a naive threshold versus the conformal detector, multi-VSA mode (MAP / BSC / HRR), a streaming twist, and a tabular fraud-style demo. This is the one to read after the quickstart.

03_sequences.py

Sequence encoding from first principles. Builds an item codebook, encodes and retrieves with the flat Sequence and the chunked HierarchicalSequence, then sweeps sequence length to show why the hierarchical variant stays near-perfect at T = 800 where flat permute-bundle collapses to about 31 % retrieval.

In progress

Files land as they are written; the numbering reserves reading order.

• 03_calibration_and_coverage.py — ECE / MCE, reliability curves, split-conformal classification and regression with coverage audits.

• 04_resonator_factorisation.py — deterministic and probabilistic resonator networks for decoding bound compositions.

• 05_real_data_eeg.py — seizure detection on a real EEG benchmark, end-to-end with calibrated probabilities.

Tutorials live under tutorials/ in the repository. The examples/ directory is a complementary cookbook: each file solves one applied problem and is meant to be read in isolation.