The optimal road trip across the U.S. according to machine learning(rhiever.github.io)
rhiever.github.io
The optimal road trip across the U.S. according to machine learning
http://rhiever.github.io/optimal-roadtrip-usa/major-landmarks.html
30 comments
Very curious about the decision metric for "major landmark" here. The San Benito County mark, for example, leads you to a dirt road in the middle of nowhere, and the major point for Delaware is just Delaware. Neither of those seem to fit a reasonable metric for a major landmark.
Seconded; I mean: I can understand why somebody who was into nuclear tourism might want to swing past the Hanford Site in Washington, but I'd bet that the majority of Americans might rather steer clear.
Hanford Site beating out the Space Needle for tourism is very interesting.
I think it's trying to pass through all 48 states, with some random landmarks and cities thrown into the mix.
So, I get what you're saying, but the particular locations are completely irrelevant to the math/CS.
If you're wondering why the route has some 'strange' segments (why get off I-10 to head down Orlando way if you're going to drive to Jacksonville later?):
The link goes to the "Major U.S. landmarks" road trip.
Other trips (including across Canada, South America, and Europe) are listed here: http://rhiever.github.io/optimal-roadtrip-usa/
The link goes to the "Major U.S. landmarks" road trip.
Other trips (including across Canada, South America, and Europe) are listed here: http://rhiever.github.io/optimal-roadtrip-usa/
When it came out a few weeks back, I noticed that it went to the South Rim of the Grand Canyon, which is the popular one. Going to the North Rim would save over 100 miles of driving. To be sure, the South Rim has the better view.
It also can't be done in winter, as its route through Yellowstone isn't open then.
It also can't be done in winter, as its route through Yellowstone isn't open then.
What does "machine learning" mean here? How does the algorithm guarantee the trip is optimal?
In general, "machine learning" means that an algorithm did it, and that the algorithm engages in some analysis of the problem space or the solution space. Usually that analysis involves an iterative or repetitive element: making several tries and modeling what makes a try good or bad, making a try and then changing the solution tiny bits to find a try that's slightly better, etc. And "good" or "bad" is determined according to a human-provided rubric (eg: +1000 points for every sight seen, -1 point for every mile driven, -10 points for every day taken, etc.)
Unless the problem is very constrained, there usually is no guarantee of optimality; here there is probably no guarantee of optimality. It might be "locally" optimal, in that there might be no better trip that differs from this one by only a tiny bit.
And, no, "machine learning" here doesn't mean the enterprise and establishment of "Machine Learning", just some algorithm that the author used.
Unless the problem is very constrained, there usually is no guarantee of optimality; here there is probably no guarantee of optimality. It might be "locally" optimal, in that there might be no better trip that differs from this one by only a tiny bit.
And, no, "machine learning" here doesn't mean the enterprise and establishment of "Machine Learning", just some algorithm that the author used.
I'm almost certain it was used incorrectly in this case. The correct term is "combinatorial optimization."
And optimal according to what criteria? Length, safety, sightseeing?
This is clearly not the optimal road trip, since it passes through Nebraska.
Also Indiana, for its eateries may now legally refuse to serve customers that use machine learning for religous reasons.
Or any state which still has the death penalty, that actually KILLS people, even teenagers in some cases, including gays.
But I guess that's not as glamorous as the latest popular cause...
But I guess that's not as glamorous as the latest popular cause...
But it does go on US2 in Montana, which is one of my criteria for an "optimal" trip through US.
But insofar as Nebraska is concerned, I have relatives there, and during one family reunion, we all went to the 4th of July parade in Hoskins. The large painted sign said "Parade at 8 am, pancake breakfast at 8:30". Not a large town.
But then I grew up 6 miles outside of a town of 3000, so I am not all that in favor of large towns.
But insofar as Nebraska is concerned, I have relatives there, and during one family reunion, we all went to the 4th of July parade in Hoskins. The large painted sign said "Parade at 8 am, pancake breakfast at 8:30". Not a large town.
But then I grew up 6 miles outside of a town of 3000, so I am not all that in favor of large towns.
Nebraska is a great place for driving really fast. One time while passing through on my way to Montana I hit 130 mph in my Integra. I don't really have anything else to say for Nebraska.
Which is a great place to drive through listening to Bruce Sprinsteen's Nebraska album (I know I did).
Ok. Following this "algorithm", California coast it is not interesting.
I pass it.
I pass it.
Yeah, and the Hanford Site in Washington is a tourist destination.
Pass!
Pass!
The cable car museum? Ok, that makes sense for SF.
http://www.cablecarmuseum.org/info.html
I'm being pedantic, but the problem being solved here is not a learning problem, but a combinatorial optimization problem.
agreed, but you're too early. The current buzzword your looking for to put in your marketing material is 'Machine Learning'. Revisit 'Combinatorial Optimization', possibly combined with 'Cloud' again in a few years.
Edit: hey great post of yours btw. http://jeremykun.com/2014/09/29/hybrid-images/
Edit: hey great post of yours btw. http://jeremykun.com/2014/09/29/hybrid-images/
Not clear to me is whether the machine learning is for the destinations, the sequencing or am I routed on scenic roads?
Skips LA - clever girl.
[1] http://www.math.uwaterloo.ca/tsp/usa50/road.html
[2] https://twitter.com/wjcook/status/575762813345480705