Here is a story from the past startup times. It's a mixture of maths and technology, the kind of things I enjoy, as an applied mathematician by education.
Sometime in 2018, I got involved in an R&D project to detect periodic events. I will skip most of the technical parts, those curious can find the entire presentation here. The presentation ends with ...
... and this is the focus of this little post.
The preable of the story is that the VP of Engineering, a lot earlier before the start of this project, "swore" that they will buy champaign to the entire Data Scienced Team if we end up using Fouries Analysis for whatever purposes. Was it for fun or they were envisioning this project coming - I have no idea.
But, the team started "assessing the field". I started looking at a simplified unimodal model. After "doodling" for a couple of hours on a one A4 size page ...
... I came with some "expressions" allowing querying the database for periodic events using ... plain SQL! Amazon Redshift was returning results within 2 minutes. I still remember our CTO saying: "You nailed it!".
However, if you check the presentation, there is no Fourier Analysis there. It's just a simple application of the basic stats, limited to the unimodal model. It was enough though to trigger the further research and this simple model was used as a benchmark.
The team, eventually, started looking for multimodal techniques and Fourier Analyses (on terabytes of data!!!) ... and the team recalled the earlier "promise" from our VP of Engineering. We definitely had a party, I don't remember the champaign though ...
No comments:
Post a Comment