groundSpring — Uncertainty Budget, Inverse Problems, Spectral Theory
Every spring's uncertainty budget — 1,164 tests, 395 validation checks, decomposes measurement error and quantifies dominant noise sources across all baseCamp papers
Domain
Sensor noise decomposition, inverse problems, sensing limits, spectral theory (Anderson localization, Almost-Mathieu operator), uncertainty quantification (jackknife, error propagation), noise floor calibration.
Repository: syntheticChemistry/groundSpring
The Science Story
groundSpring is the measurement spring — the spring that answers “how much of your signal is real and how much is noise?” Every other spring depends on it. When airSpring reports R²=0.97, groundSpring decomposes the 3% residual into humidity sensor noise (66%), wind measurement error (21%), and radiation uncertainty (13%). When wetSpring reports 5,000-read saturation, groundSpring quantifies the noise floor. When hotSpring reports 0.000% energy drift, groundSpring provides the jackknife confidence interval.
The core insight: uncertainty is not a footnote — it is the signal. Bazavov’s jackknife for lattice QCD, Anderson’s localization for spectral theory, and FAO-56’s humidity correction are all instances of the same pattern: decompose the error budget, find the dominant term, reduce it.
Headline Results
- 1,164 tests across 3 crates, 0 failed
- 395/395 validation checks (340 core + 55 NUCLEUS)
- 29/29 Python baselines with math parity proven
- 110 barraCuda delegations (67 CPU + 43 GPU)
- guideStone Level 3 — bare + IPC wired
- Contributes to every baseCamp paper via uncertainty quantification
Validation Phases
| Phase | Key Result |
|---|---|
| Decomposition | Signal vs noise separation across 10 scientific domains |
| Spectral | Anderson localization, Almost-Mathieu operator, transport exponents, band edge analysis |
| Jackknife | Bazavov QCD jackknife, error propagation, noise floor calibration |
| Cross-Spring | airSpring humidity 66% of ET₀ uncertainty; wetSpring 5,000-read saturation; neuralSpring sensor noise floors |
| GPU | 43 GPU-delegated operations via barraCuda — inverse problems on consumer GPUs |
Researchers Reproduced
| Researcher | Department | Domain |
|---|---|---|
| Alexei Bazavov | CMSE + Physics, MSU | Lattice QCD jackknife, autocorrelation |
| Ilya Kachkovskiy | Math, MSU | Anderson localization, spectral theory |
| Younsuk Dong | BAE, MSU | ET₀ uncertainty decomposition |
What the Constraint Revealed
Making uncertainty a first-class citizen (not an afterthought) forced every spring to declare its noise model. This created a natural calibration cascade: groundSpring validates the measurement, the spring uses the measurement, and the provenance chain records both. The constraint also drove the GPU uncertainty pipeline — inverse problems on consumer GPUs via barraCuda, where the GPU speedup matters most for Monte Carlo error estimation.
Cross-Spring Connections
- → airSpring: “humidity dominates ET₀ uncertainty at 66%” — the uncertainty budget shaped irrigation engineering
- → wetSpring: Sequencing noise calibrates rarefaction; 86 named tolerances with provenance
- → hotSpring: Spectral primitives + QCD inverse problems; jackknife for lattice observables
- → neuralSpring: Sensor noise floors for ESN/LSTM training data validation
- → lithoSpore: B1-B4 statistical methods — model fitting, fixation probability, AIC/BIC model selection for LTEE modules
Notebooks (5)
| # | Notebook | Focus |
|---|---|---|
| 01 | Noise Decomposition | Sensor noise, temporal drift, spatial variability |
| 02 | Spectral Analysis | Anderson localization, transport exponents |
| 03 | Jackknife & Uncertainty | Bazavov QCD jackknife, FAO-56 error propagation |
| 04 | GPU Inverse Problems | Consumer GPU Monte Carlo, barraCuda delegation |
| 05 | Cross-Spring Budget | How groundSpring uncertainty flows into every other spring |
baseCamp Papers
Papers 01, 02, 03, 04, 05, 06, 07, 10, 12, 16 — see baseCamp Science for full list.
groundSpring contributes uncertainty quantification to every baseCamp paper. The papers listed are those where groundSpring methods are directly cited.