SiPANN 1.1.1.post1

Artifical Neural Networks for use with Silicon Photonics

Silicon Photonics with Artificial Neural Networks. SiPANN aims to implement various silicon photonics simulators based on machine learning techniques found in literature. The majority of these techniques are linear regression or neural networks. As a results SiPANN can return scattering parameters of (but not limited to)

• Half Rings

• Arbitrarily shaped directional couplers

• Racetrack Resonators

• Waveguides

And with the help of simphony and SiPANN’s accompanying simphony wrapper

• Ring Resonators

• Doubly Coupled Rings

• Hybrid Devices (ie Green Machine)

Installation

SiPANN is distributed on PyPI and can be installed with pip:

pip install SiPANN

Developmental Build

If you want a developmental build, it can be had by executing

git clone https://github.com/contagon/SiPANN.git
pip install -e SiPANN/

This development version allows you to make changes to this code directly (or pull changes from GitHub) without having to reinstall SiPANN each time.

You should then be able to run the examples and tutorials in the examples folder, and call SiPANN from any other python file.

Note

If installing on Windows, one of SiPANN’s dependencies, gdspy, requires a C compiler for installation. This can be bypassed by first installing the gdspy wheel. This is done by downloading the wheel from gdspy, navigating to the location of the wheel, and executing

pip install gds*.whl

After this simply install SiPANN using your desired method.

References

SiPANN is based on a variety of methods found in various papers, including:

[1] A. Hammond, E. Potokar, and R. Camacho, “Accelerating silicon photonic parameter extraction using artificial neural networks,” OSA Continuum 2, 1964-1973 (2019).

Bibtex citation

@misc{SiP-ANN_2019,
        title={SiP-ANN},
        author={Easton Potokar, Alec M. Hammond, Ryan M. Camacho},
        year={2019},
        publisher={GitHub},
        howpublished={{https://github.com/contagon/SiP-ANN}}
}