Music is Math

Posted 2010.09.11

I have been pretty obsessed with a little side project over the last couple of weeks which explains why activity on the left column on this blog has been low … . It all started about a month ago when a colleague in London told me about the Korg drum computer app on the iPad which transforms the iPad into a wicked beat machine. I was immediately hooked. After a couple of days of beat making I decided I was ready for something bigger and I started looking into other music software. That’s when a very kind Best Buy employee introduced me to Reason. My life will never be the same again!

I have no musical background, have never played an instrument and never went to music school. I was a complete novice when I picked up Reason 3 weeks ago. A couple of Youtube reason tutorials later, I launched my own little one man band on myspace.

The software is incredible. It allows you to make any sound imaginable, from horns to electric guitars, from techno synthesizers to monk chants and crickets. These instruments play notes on different tracks which you can then combine in a mixer. The picture below shows a Reason screenshot with a song that is played on 12 different tracks.

It also has a number of drum computers you can use to make your own beats or you can select from a library of hundreds of pre made loops. The possibilities seem endless and while you can make things extremely complicated (Reason is used by pro’s like the Prodigy), I found myself finishing my first song in just 2 days.

What quickly became clear in this process is how similar making music is to analytics. While I was composing and arranging, it felt like I was using the part of my brain I use while writing code in SAS. That shouldn’t be surprising. The theory about the similarities between music and mathematics goes back all the way to Plato. There is even a wikipedia entry devoted to the subject. The screenshot above shows has the structure of a database. Every track is a file of data (notes) that is linked through the rhythm and the beat.

Analytical skills came in really handy in interpreting the notes, especially when you have never had musical training. I quickly discovered how useful midi files are. Midi files are similar to mp3 files but rather than containing the sound waves, they contain the musical instructions for the instruments (the notes, the rhythm, the velocity and length of the notes, …). You can read them automatically into Reason and they show up like different tracks on a screen like the one shown above. You can then select which instruments you would like to play which tracks and then move blocks of notes around to compose and arrange different variations of the song. A lot of midi files can be downloaded from the web.

Not every song is available on midi though. When I wanted to make a dance version of one of my favorite 90’s grunge songs I couldn’t find the midi for the song anywhere. I was stuck. So I started looking for a solution and came across Widisoft, a software package that reads in sound files and translates them to midi files. Here is a screenshot of my grunge song translated by widisoft. It is remarkable how similar it looks to the punchcards that came out of the first computers. It’s data in it’s purest form!

At first sight the screenshot looks like random dots on a couple of bars. But when you listen to the song and look at the data at the same time you quickly start to see the patterns that identify which data points correspond to which sequences and instruments. Some intuitive pattern recognition allows you to split this data stream into different organized tracks pretty easily. This creation of order out of data chaos is very similar to what data analysts do when they analyze large data sets.

I decomposed the data, mapped the different tracks to new instruments, put it all on a house beat, rearranged the sequences and recorded Slaphappy. Fans of early 90’s Seattle grunge might recognize some of the riffs.