gpCAM
[!NOTE] Upgrading from 8.3.x? gpCAM 8.4 targets
fvgp ~= 4.8and renames a few constructor kwargs (gp2Scale_dask_client → dask_client,gp2Scale_linalg_mode → linalg_mode,gp_rank_n_update → rank_n_update), and replacescalc_inv=Truewithlinalg_mode="CholInv". Full migration notes are in HISTORY.rst.
gpCAM (gpcam.lbl.gov) is an API and software designed to make advanced Gaussian Process function approximation and autonomous data acquisition/Bayesian Optimization for experiments and simulations more accurate, faster, simpler, and more widely available. The tool is based on a flexible and powerful Gaussian process regression at the core. The flexibility stems from the modular design of gpCAM which allows the user to implement and import their own Python functions to customize and control almost every aspect of the software. That makes it possible to easily tune the algorithm to account for various kinds of physics and other domain knowledge and to identify and find interesting features, in Euclidean and non-Euclidean spaces. A specialized function optimizer in gpCAM can take advantage of HPC architectures for fast analysis time and reactive autonomous data acquisition. gpCAM broke a 2019 record for the largest exact GP ever run! Below you can see a simple example of how to set up an autonomous experimentation loop.
Usage
The following demonstrates a simple usage of the gpCAM API (see interactive demo).
!pip install gpcam
from gpCAM import GPOptimizer
my_gp = GPOptimizer(x_data,y_data,)
my_gp.train()
train_at = [10,20,30] #optional
for i in range(100):
new = my_gp.ask(np.array([[0.,1.]]))["x"]
my_gp.tell(new, f1(new).reshape(len(new)))
if i in train_at: my_gp.train()
Designing experiments with Claude Code
gpCAM ships with a set of Claude Code skills that guide an AI assistant through designing autonomous experiments — custom kernels, acquisition functions, noise models, and the full ask/tell/train loop. Experimentalists who want smart, autonomous data acquisition without deep knowledge of GP math or the gpCAM API can use these skills to design autonomous experiments.
Installing the gpCAM marketplace in Claude Code
The repo ships as a Claude Code plugin marketplace. Inside any Claude Code session, run:
/plugin marketplace add lbl-camera/gpCAM
/plugin install gpcam@gpcam
The first command registers this repo as a marketplace; the second installs the gpcam plugin from it, which bundles all of the skills below. After install, the skills are available to Claude in any project — no need to clone the repo locally.
To update later, run /plugin marketplace update gpcam; to remove, /plugin uninstall gpcam@gpcam.
Available skills
| Skill | Description |
|---|---|
| experiment-designer | End-to-end autonomous experiment design. Translates a scientist's description of their measurement into a complete gpCAM script. |
| kernel-designer | Design and compose custom kernel functions that encode domain knowledge (smoothness, periodicity, symmetry, anisotropy). |
| acquisition-functions | Write custom acquisition functions that encode experimental priorities (exploration vs exploitation, multi-objective, constraints). |
| prior-mean-functions | Encode known physics or expected trends as prior mean functions. |
| noise-functions | Model position-dependent or heteroscedastic noise from detector characteristics. |
| cost-functions | Account for motor travel time, settling, directional costs, and zone-based penalties. |
| uncertainty-calibration | Validate that the posterior's error bars are trustworthy — scoring rules (NLPD, CRPS, MSLL), coverage/calibration curves, and what to do when the model is over- or under-confident. |
| troubleshooting | Map a gpCAM error message to its cause and fix — non-positive-definite covariance, bounds mismatches, NaNs, Dask hangs, pickling, and silent failures. |
| gp2scale-advanced | Large-scale experiments (>10k points) using sparse kernels and Dask distributed computing. |
| multi-task-advanced | Multi-output / function-valued experiments with fvGPOptimizer. |
| transformed-optimizers-advanced | Constrained observations: LogGPOptimizer for y > 0 and LogitGPOptimizer for y ∈ [0, 1] — predictions and credible intervals stay inside the constrained range. |
Once installed, the skills activate automatically when you describe an experiment design problem to Claude, or you can invoke one explicitly (e.g. "use the experiment-designer skill to set up an adaptive XRD scan").
Other agentic platforms
The skills are also compatible with any harness that reads SKILL.md files (e.g. OpenClaw) — clone the repo and point your assistant at the skills/ directory. When this repo is present in your working directory, Claude Code also picks up the root CLAUDE.md and skills/ directory automatically, so the marketplace install is only needed for use outside the repo.
Credits
Main Developer: Marcus Noack (MarcusNoack@lbl.gov)
This code was developed with help from Ron Pandolfi (LBNL), Mark Risser (LBNL), Hengrui Luo (Rice U.), and Vardaan Tekriwal (UCB).
Additional contributions and insights came from across the community, in particular, Kevin Yager, Masafumi Fukuto, and their teams (Brookhaven National Lab).
We acknowledge support from several DOE ASCR, BER, and BES projects, including CAMERA (James Sethian, Jeff Donatelli), SPECTRA (Sherry Li), and CASCADE (Bill Collins), as well as support directly from Lawrence Berkeley National Laboratory.
This package uses the HGDL package of David Perryman and Marcus Noack, which is based on the HGDN algorithm by Noack and Funke.
2 maintainers
Release files for gpcam 8.5.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| gpcam-8.5.0.tar.gz | 147.3 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| gpcam-8.5.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 199.9 kB
Release files / gpcam-8.5.0.tar.gz
| Download URL | gpcam-8.5.0.tar.gz |
|---|---|
| Size | 147.3 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
d055e6458f3276dfaea7b93b0e7bef608e4a6d55c67028b2f9806439136e6754
|
|
BLAKE2b-256 checksum How to use checksums |
bb14d000556eb0a2505392a6438763205202a757e8f3737009d39cc227954106
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Release files / gpcam-8.5.0-py3-none-any.whl
| Download URL | gpcam-8.5.0-py3-none-any.whl |
|---|---|
| Size | 52.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
f4ffeb57504f683bc2c921f2486f8b833b0494a3c20b15c7649d3f2391b21e10
|
|
BLAKE2b-256 checksum How to use checksums |
fff2c4acf28901710b9442688496cfb4de1fc98e2ff6ea3c535967ee86040949
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
