Common Framework for Inference (CoFI)

The Common Framework For Inference (CoFI) project aims to revolutionise inversion capabilities in geophysics by developing an open and adaptable software platform that facilitates collaboration and innovation in understanding the subsurface.

Overview

CoFI, led by Professor Malcolm Sambridge from The ANU, is a collaborative effort between The ANU, CSIRO’s Deep Earth Imaging Future Science Platform, and the University of Durham. This project will develop a flexible high-level programming language for inversion, connecting researchers with leading-edge algorithms and educators with interactive teaching resources.

“Typically, geoscience researchers who develop inversion methods need problems to test them on, and those who need to solve inversion problems need leading-edge algorithms to solve them. The idea behind CoFI is to allow the two to meet in the middle while providing an ease-of-use platform for educators.

We are excited to be embarking on this new project, the first of its kind in the world, with our partners in AuScope and CSIRO." - Professor Malcolm Sambridge, Project Lead

The challenge

Existing tools for geoscience inference problems often lack flexibility and ease of integration with various datasets and algorithms, limiting their effectiveness and accessibility. This project aims to address these challenges by creating a unified, open-source platform that simplifies and enhances the development and application of geoscience inference solutions.

The outcomes

  • Designed and implemented the CoFi computational software architecture.

  • Released version 1.0 of the CoFi software platform to the scientific community.

  • Created a test suite of standalone inversion problems with working examples.

  • Developed a continuous improvement protocol and user code insertion hooks.

  • Established a collaborative framework for researchers to deploy their data and models.

What are the benefits?

  • Streamlined research: The CoFI infrastructure helped simplify solving inverse problems through a unified framework that connects various data sets and forward solvers, eliminating the time for researchers to develop custom algorithms.

  • Enhanced reproducibility: researchers using CoFI can ensure their methods and results are more reproducible, which is critical to fostering scientific understanding.

  • Greater Accessibility & Integration: The software infrastructure helps link with other third-party packages like SciPy and PyTorch to tackle a wide range of inverse problems. It also offers an extensive suite of tutorials and examples to make the framework accessible to everyone.

Who will benefit?

Researchers, government agencies, and industry professionals involved in geoscience, hydrogeology, and related fields will benefit from the flexible and powerful CoFi platform. This project will enhance their ability to develop and apply advanced inference solutions, promoting innovation and collaboration.

Access

  • Data Access: Access to the CoFi software platform will be provided through open-source mechanisms such as GitHub.

  • Tools and Software Access: The software will be available via utility install protocols like Unix pip, with usage and demonstrations facilitated through Python, IDEs, Jupyter notebooks, and Unix OS.

  • Workshops and Training: Training workshops will support users in utilising the CoFi platform effectively.

Acknowledging AuScope

This project was made possible by support from the National Collaborative Research Infrastructure Strategy (NCRIS) through AuScope. Acknowledging AuScope and NCRIS helps us demonstrate the value of shared research infrastructure, ensuring continued support and resources for the research community.

If you helped deliver this project or have benefited from its outputs, please credit AuScope so we can include your work in our impact reporting. For examples of acknowledgment, please visit our ‘How to Acknowledge AuScope’ page.

We’d love to see your work—please tag us on social media using:

@auscope | #AuScopeImpact | #NCRISimpact

 
 

Project Name
Common Framework For Inference project (CoFI)

Project Lead

Timeframe
2022 to 2023

Status
Complete

Funding
AuScope Pilots 3

Hosts
Australian National University (ANU)
CSIRO

NCRIS Collaborators
Pawsey
National Computational Infrastructure (NCI)

Other Collaborators
Durham University

AuScope Programs

SAMAuScopePilot 3, SAM, CoFI, Inference, Forward Solvers, Data Modelling