Metadata-Version: 2.4
Name: dgpsi
Version: 2.6.0
Summary: Deep and Linked Gaussian Process Emulations using Stochastic Imputation
Author-email: Deyu Ming <deyu.ming.16@ucl.ac.uk>
License: MIT
Project-URL: Homepage, https://github.com/mingdeyu/DGP
Keywords: surrogate modelling,deep learning,stochastic EM,elliptical slice sampling
Requires-Python: <3.13,>=3.9
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy>=1.18.2
Requires-Dist: numba>=0.51.2
Requires-Dist: matplotlib>=3.2.1
Requires-Dist: tqdm>=4.50.2
Requires-Dist: scipy>=1.4.1
Requires-Dist: scikit-learn>=0.22.0
Requires-Dist: jupyter>=1.0.0
Requires-Dist: dill>=0.3.2
Requires-Dist: pathos>=0.2.9
Requires-Dist: multiprocess>=0.70.13
Requires-Dist: psutil>=5.8.0
Requires-Dist: tabulate>=0.8.7
Provides-Extra: interactive
Requires-Dist: jupyter; extra == "interactive"
Dynamic: license-file

# dgpsi
[![GitHub release (latest by date including pre-releases)](https://img.shields.io/github/v/release/mingdeyu/DGP?display_name=release&include_prereleases&style=flat-square)](https://github.com/mingdeyu/DGP/releases)
[![Conda](https://img.shields.io/conda/dn/conda-forge/dgpsi?label=Conda%20Downloads&style=flat-square)](https://anaconda.org/conda-forge/dgpsi)
![Conda](https://img.shields.io/conda/pn/conda-forge/dgpsi?color=orange&style=flat-square)
[![Read the Docs (version)](https://img.shields.io/readthedocs/dgpsi/latest?style=flat-square)](https://dgpsi.readthedocs.io)
[![GitHub R package version](https://img.shields.io/github/r-package/v/mingdeyu/dgpsi-R)](https://github.com/mingdeyu/dgpsi-R)

## For R users
The `R` interface to the package is available at [`dgpsi-R`](https://github.com/mingdeyu/dgpsi-R).

## A Python package for deep and linked Gaussian process emulations using stochastic imputation (SI)
`dgpsi` currently implements:

* Gaussian process emulations with separable or non-separable squared exponential and Mat&eacute;rn-2.5 kernels.
* Deep Gaussian process emulation with flexible architecture construction: 
    - multiple layers;
    - multiple GP nodes;
    - separable or non-separable squared exponential and Mat&eacute;rn2.5 kernels;
    - global input connections;
    - non-Gaussian likelihoods (Poisson, Negative-Binomial, heteroskedastic Gaussian, and categorical);
* Linked emulation of feed-forward systems of computer models:
    - linking GP emulators of deterministic individual computer models;
    - linking GP and DGP emulators of deterministic individual computer models;
* Multi-core predictions from GP, DGP, and Linked (D)GP emulators;
* Fast Leave-One-Out (LOO) cross validations for GP and DGP emulators.
* Calculations of ALM, MICE, and VIGF sequential design criterions.
* ![Feature Badge](https://img.shields.io/badge/Feature-New-orange) [Large-scale GP, DGP, and Linked (D)GP emulations.](https://github.com/mingdeyu/DGP/blob/master/demo/vecchia_SI.ipynb)
* ![Feature Badge](https://img.shields.io/badge/Feature-New-orange) [Scalable DGP classification using Stochastic Imputation.](https://github.com/mingdeyu/DGP/blob/master/demo/DGP_classification.ipynb)

## Installation
### Release version
The current stable release of `dgpsi` supports Python versions 3.9, 3.10, 3.11, and 3.12. The stable release can be installed via `pip`:

```bash
pip install dgpsi
```

or `conda`:

```bash
conda install -c conda-forge dgpsi
```

However, to achieve the best computational performance of the package (e.g., on Apple Silicon), we recommend the following steps for the installation:
* Download and install `Miniforge3` that is compatible to your system from [here](https://github.com/conda-forge/miniforge).
* Run the following command in your terminal app to create a virtual environment called `dgp_si`:

    ```bash
    conda create -n dgp_si python=3.9.13 
    ```

* Activate and enter the Conda environment:

    ```bash
    conda activate dgp_si
    ```

* Install `dgpsi`:
    - for Apple Silicon users, you could gain speed-up by switching to Apple's Accelerate framework:

        ```bash
        # for macOS <= 13.2
        conda install dgpsi "libblas=*=*accelerate"

        # for macOS >= 13.3
        conda install dgpsi "libblas=*=*newaccelerate"
        ```

    - for Intel users, you could gain speed-up by switching to MKL:

        ```bash
        conda install dgpsi "libblas=*=*mkl"
        ```

    - otherwise, simply run:

        ```bash
        conda install dgpsi
        ```

### Development version
If you want to try the newest features and fixes before the next release, you can install the development build using the provided Conda environment YAMLs, which select the optimal BLAS and dependencies for your platform.

* Clone the repository:

    ```bash
    git clone https://github.com/mingdeyu/DGP.git
    cd DGP
    ```

* Pick the right environment file for your platform:

    | Hardware / Platform                      | BLAS backend   | YAML file                       |
    | ---------------------------------------- | -------------- | --------------------------------|
    | Apple Silicon (macOS <= 13.2)            | Accelerate     | `env-arm64-accelerate.yaml`     |
    | Apple Silicon (macOS >= 13.3)            | New Accelerate | `env-arm64-newaccelerate.yaml`  |
    | Intel CPU (macOS/Linux/Windows)          | MKL            | `env-intel-mkl.yaml`            |
    | Other (Linux/Windows)                    | OpenBLAS       | `env-other-openblas.yaml`       |

* Create and activate the Conda environment:

    ```bash
    # replace the yaml filename with the one for your platform
    conda env create -f env-arm64-accelerate.yaml
    conda activate dgp_si_dev
    ```
    > **Tip:** You can override the Conda environment name by appending `-n <myenv>` to the create command.

* Install the dev version from your local clone:

    ```bash
    pip install --no-deps --no-build-isolation .
    ```

## Demo and documentation
Please see [demo](https://github.com/mingdeyu/DGP/tree/master/demo) for some illustrative examples of the method. The API reference 
of the package can be accessed from [https://dgpsi.readthedocs.io](https://dgpsi.readthedocs.io).

## Tips
* Since SI is a stochastic inference, in case of unsatisfactory results, you may want to try to restart the training multiple times even with initial values of hyperparameters unchanged;
* The recommended DGP structure is a two-layered one with the number of GP nodes in the first layer equal to the number of input dimensions (i.e., number of input columns) and the number of GP nodes in the second layer equal to the number of output dimensions (i.e., number of output columns) or the number of parameters in the specified likelihood.

## Contact
Please feel free to email me with any questions and feedbacks: 

Deyu Ming <[deyu.ming.16@ucl.ac.uk](mailto:deyu.ming.16@ucl.ac.uk)>.

## Research Notice
This package is part of an ongoing research initiative. For detailed information about the research aspects and guidelines for use, please refer to our [Research Notice](./RESEARCH-NOTICE.md).

## References
> [Ming, D. and Williamson, D. (2023) Linked deep Gaussian process emulation for model networks. arXiv:2306.01212.](https://arxiv.org/abs/2306.01212)

> [Ming, D., Williamson, D., and Guillas, S. (2023) Deep Gaussian process emulation using stochastic imputation. <i>Technometrics</i>. 65(2), 150-161.](https://doi.org/10.1080/00401706.2022.2124311)

> [Ming, D. and Guillas, S. (2021) Linked Gaussian process emulation for systems of computer models using Mat&eacute;rn kernels and adaptive design, <i>SIAM/ASA Journal on Uncertainty Quantification</i>. 9(4), 1615-1642.](https://epubs.siam.org/doi/abs/10.1137/20M1323771)
