Automatic differentiation - an RSE's eye view

Add to your list(s) Download to your calendar using vCal

• Matt Graham - ARC, UCL
• Thursday 20 November 2025, 13:00-14:00
• Room A, West Hub.

If you have a question about this talk, please contact Jack Atkinson.

Automatic or algorithmic differentiation (AD) underpins much of the current boom in use of machine learning methods but is also widely used in other scientific computing contexts.

In this talk I will give an overview of what automatic differentiation is and a brief summary of its history. I will then review how it relates to symbolic and numerical differentiation, how forward- and reverse-mode AD differ and some of the different approaches to implementing AD frameworks, before demonstrating how AD is used in practice with some applied examples. I will conclude with some discussion of my experiences of using various automatic differentiation implementations in research software projects I have worked on, particularly from a context of the trade-offs between ease of use and maintainability and generality of code a framework can differentiate.

This talk is part of the RSE Seminars series.

This talk is included in these lists:

• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge talks
• Chris Davis' list
• Interested Talks
• RSE Seminars
• Room A, West Hub
• Trust & Technology Initiative - interesting events
• bld31
• ndk22's list
• ob366-ai4er
• rp587
• se393's list
• yk449

Note that ex-directory lists are not shown.