The choice between exploring and exploiting
The road less traveled or the more familiar: How do you choose?
When you point your wheel at the road you’ve got a choice: Do you take an old familiar route or try something new? In machine learning it’s usually phrased as a choice between exploring and exploiting, looking around for higher returns or maximizing a sure thing. Do you pick a new road that might have better views, a smoother surface, fewer loose dogs? Or do you fall back to the familiar, where you know every hill — and maybe even the dogs’ names?
Paul and I planned a long ride east of town, but when I drew up the route I picked a couple of roads I’d never been on: Old Beavers Road and Piney Woods Road. I’ll confess: I just liked the sound of Old Beavers, but I chose Piney Woods to keep us off a relatively busy stretch of road. It was a misty morning, and when we turned onto Old Beavers the road unrolled before us through horse farms and gentle hills. The road was a little rough, but there was almost no traffic. Definitely one to keep on the list.
A little further down we turned onto Piney Woods. Mobile homes, dogs running loose or chained up and barking, and short, punchy climbs that were just long enough to dissolve any momentum we carried. No need to revisit that one.
During the week, we probably tend to exploit the roads we know well. We know what the traffic will be like, how long it will take to get home, and all of those other details you learn ride after ride, week after week. Time is the key variable, the limit on whether we exploit or explore.
When it’s one ride on one day, we know what that time limit is. Usually it’s what our schedule will allow.
But in the larger picture, and ultimately — in its grimmest and most literal sense — we don’t know what our time limit is. We have no idea how much time we’ll have to explore new roads and how much time we have to ride the routes we love. So we have to mix it up. I rode Old Beavers again this weekend, and skipped Piney Woods. And I found another road in that area that I’ll throw into the mix some time. Soon. It looks pretty promising from the Google street view.
Unlike a business strategy where you can calculate a return on your investment over a set period of time, this is a decision with only one of two results over an unknown period of time. I’ll either have a new road to return to, or I will have learned something about a road I won’t ever have to travel again. Just try to calculate a return on that.
The ride: Dam Loop with Paul