The first time I came across Steve Bruntons YouTube channel I came up with all these excuses for not to watch: “This is too advanced math for me”, “Math at this level is not really useful for practical application” etc. Now he has become one of my biggest influencers 🙂

I am not sure which video that triggered my interest for “EigenSteve” and his colleagues at Washington University, but I think it was his “Data-Driven Science and Engineering”-video series. I realized that Steve Brunton has a special ability to make math both really interesting, understandable and useful. His enthusiasm and practical examples of use within dynamics, system identification, control theory, machine learning, fluid dynamics, digital image processing and more makes every video a treat to watch and very transferrable to many projects I am working on.

What makes his videos especially useful are the sharing of code for matlab and python. When trying to understand a subject with cutting-edge research you allways have to look into the papers written to get the details. In better cases you can get some code snippets or links to some github page that is difficult to implement and use. In worse cases you have to link together small pieces and you rarely end up with the same code that is described in the paper.

Steve Brunton, Nathan Kutz and their colleagues is flipping this upside down by making many complete books available for free, sharing all their code and papers publicly, making videos describing in detail the math in an easy way and making walkthroughs videos of their code. This makes all of their tools very available for use in any modern engineering application.

Many of the tools that were barely or not known to me before starting to watch his videos. Now I have watched all of his videos and I am currently using or have used techniques such as DMD, SINDy, SVD/PCA, sparse sensing, HAVOK, LQG and MPC.

One very central technique to many of the other techniques is Singular Value Decomposition (SVD) and it is a technique that should come to mind when working with many different subjects:

I would also like to highlight SINDy – The toolbox for making white-box parsimonious, equation based models directly from data. It seems like every video coming out from Steve Bruntons group is making this technique better in every way by being more robust, requiring less data, working faster etc.

I feel we all owe Steve Brunton a big thank you – Your work has made advanced and very useful tools understandable and available for the whole world to use.