gp_foundations

gp_foundations is a research repository built from a curated PDF corpus on Gaussian processes, Bayesian optimization, numerical linear algebra, multi-output inference, non-stationary modelling, two-timescale updates, Kalman filtering, chaos, and concurrency.

The repository is package-first: the Python toolkit under src/gp_foundations/ is the main executable surface, while docs/ and notebooks/ mirror the source papers in a repo-native form.

Status

This is an early-stage research toolkit. The repository is structured and tested, but the priority is coherent scientific code and documentation rather than release packaging.

Scope

Phase one implements:

• Gaussian process regression with marginal likelihood evaluation
• Opt-in hyperparameter learning for exact Matérn GP and multi-output GP models
• Cholesky-based numerical helpers
• Bayesian optimization acquisition functions
• Intrinsic coregionalization for multi-output Gaussian processes
• Practical non-stationary kernels and a two-timescale update scheduler
• Offline WiFi replay and simulation utilities built on the shared MOGP stack
• Supporting Kalman, chaos, and concurrency utilities

Layout

• src/gp_foundations/: package code
• src/gp_foundations/wifi_research/: offline replay and simulation layer
• benchmarks/: fixed-vs-learned benchmarking scripts
• tests/: unit and smoke tests
• docs/: curated Markdown documents derived from the PDFs
• notebooks/: notebook walkthroughs that mirror the documents
• references/pdfs/: original source PDFs copied into the repository

Local development

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -e .[dev,plotting]
python -m pytest

If pytest is not installed locally, the test suite can still run via the standard library:

PYTHONPATH=src python3 -m unittest discover -s tests -v

Example

import numpy as np
from gp_foundations.gp import GaussianProcessRegressor
from gp_foundations.kernels import MaternKernel

X = np.linspace(0.0, 1.0, 8)[:, None]
y = np.sin(2.0 * np.pi * X[:, 0])

gp = GaussianProcessRegressor(MaternKernel(length_scale=0.2), noise=1e-4)
gp.fit(X, y)
posterior = gp.posterior(np.linspace(0.0, 1.0, 100)[:, None])

License

This repository is released under the MIT License. See LICENSE.