A letter to my favorite algorithm
Dear random forest, my name’s Juan, a huge fan.
Lately, I’ve been thinking a lot about you, wondering how you are feeling, and pondering about what the future has for you. You see, I’m sad that the sound of your leaves rustling is not being heard the way it used to. Since some years ago, your other fans and practitioners decided to leave the trails of your forest in exchange for a more neural and networked life. And while I’m sure they had their reasons, this makes me sad. Let me tell you something.
When I started to learn about machine learning, I had difficulties understanding many of the essential concepts that are part of our beloved field. Oh, my friend, yes. You have no idea. However, my luck turned when I met your youngest sibling, the decision tree, who I call Detri. Back then, through Detri, I learned the bases and essential concepts I so desperately needed. I have no idea why, but when I started to see things in terms of branches, leaves, and flowcharts I had an enlightenment moment and suddenly many of the topics that were a struggled became more tolerable to comprehend. How happy I was!
See, right after this realization, I started using decision trees for many of my pet projects; Pokemon data, gaming data, and the inevitable Iris dataset were some of the ones who traversed the twigs and boughs of your leafy brother. Not only that, for my thesis, guess what I used? Exactly.
During one of these tree-climbing expeditions, I decided to become a dendrologist and learn how such an algorithm is structured. So, I looked to answer some of the following questions: What exactly is a node? How is the tree made? Oh, wait, can I also use it for regression? I was purely amazed. Amid this adventure, I discovered you. You random forest, an ensemble of my fond trees. What a good day that was.
After I got to know you, I went to many, many hikes through your forest. Suddenly, I had updated most of my previous models, to you — end the era of the lonely tree, and welcome to the ages of the multiple ones! And well, I don’t want to brag, but I knew several things about you; I went from a dendrologist to a forester. Want to know something great? Back when I got my first job, I used you for my first supervised learning task! Of course, I did! I remember fitting a spring-ish, vast, but-not-so-deep, forest that looked like a National Park. And wow, how graceful you were. Waves and waves of spam, you controlled (and with high confidence, by the way!), making me (and my bosses) really satisfied.
Unfortunately, after some time, we had to replace you. You just became a bit too heavy and slow to train. I don’t mean to offend you! It’s just that our requirements changed, and so did our strategy… or maybe we were just not ready for your grandeur. Even after you were gone, I kept finding traces of your twigs, leaves, and branches among my own forest of code, and every single time it brought a smile to my face; “aha! Looks who’s here”, I said when I saw parts of your whole in line 37. Gone, but not forgotten.
And similarly to how I stopped using you at work, it seems like most of the world also did. Even though Google Trends’ data says otherwise, I feel like your rustling is not as loud as before. Now, I seldom see use cases ruled by you, nor people speaking about you in conferences; it’s like autumn finally arrived at the woods.
But hey, I also heard the good stories! Lately, your cousin gradient boosted trees or Boostry, as I call it, has been all over the news, and that’s fantastic. I mean, with you as a mentor, what could have gone wrong.
So, yes, my good friend, those were my words. I sincerely hope everything is okay with you. However, to end in a positive note, I’ll say that if there’s something I have learned, is that in this field, spring is always around the corner, so I’m sure I’ll be hearing from you.
Your fan, Juan.