> 26.8% fewer search steps than standard A∗ search
So you're saying I shouldn't read this paper?
A* is not particularly good. If all the algorithm can do is mirror it by training on its output what's the point?
So you're saying I shouldn't read this paper?
A* is not particularly good. If all the algorithm can do is mirror it by training on its output what's the point?
I publish in this area and this is a common thing for reviewers to bring up when authors don't report wall clock time. And then for papers to be rejected. What's the value in making an algorithm that's drastically slower? Not much.
Perhaps as an important stepping stone? Deferred optimization and all that.
Some people call the flag planting. You publish something that doesn't work with keywords you suspect will be important in the future. Then hope others will cite you.