amitp

18 ms·

151.
by amitp 8y ago
Also check out A* if you want to go 1000 times faster than JPS ;------) A* does well when it is given a graph of all the decision points that matter. An unweighted grid is full of locations where it doesn't matter — e.g. whether you go
152.
by amitp 8y ago
Glad to hear it! :)
153.
by amitp 8y ago
Thank you! I'm working on a new map project using some of those techniques.
154.
by amitp 8y ago
It wasn't long ago that Game AI Pro said “bidirectional pathfinding for A* is usually a poor choice”[1]. There's been some recent work presented at GDC 2018 about bidirectional search[2] but even there they said it was only useful
155.
by amitp 8y ago
Also check out the interactive diagrams for the graph search chapter of that book [1]. I especially liked the "shoes of a search agent" one, which emphasizes that the algorithm doesn't "see" the whole graph like a h
156.
by amitp 8y ago
Cool! When I read Hawkins's book _On Intelligence_ back in the mid-2000s I had thought it'd be cool to generate music by having the system predict what was novel and what was familiar. By mixing some novel and familiar notes/
157.
by amitp 8y ago
I agree, Voronoi diagrams are so much fun! For that map project, at the last minute I switched to something slightly different. The page still says "Voronoi" but the actual demo doesn't use Voronoi. The Voronoi diagram uses t
158.
by amitp 9y ago
Alternate format: https://threadreaderapp.com/thread/980968298002006016.html
159.
by amitp 9y ago
This is my favorite podcast of all time. It's completely fascinating and yet, as with other history podcasts, it puts me to sleep in 5 minutes. I listen to them over and over to catch bits in those 5 minutes. You get to learn thing lik
160.
by amitp 9y ago
If your goal is to make it "cellular" you can get an even more cellular appearance by using the Delaunay triangle centroids instead of the circumcenters like Voronoi diagrams use. Side by side comparisons https://www.
161.
by amitp 9y ago
I wanted to thank you not only for this new page but also for your blog and distill.pub, especially the research debt page!
162.
by amitp 9y ago
That was one of the five sites I found for eyeglass materials. These are the others: - http://www.allaboutvision.com/lenses/how-to-choose.htm - https://web.archive.org/web/20090815102701/http:
163.
by amitp 9y ago
The javascript data structures and algorithms are on my new github page: - https://github.com/redblobgames/dual-mesh/ - https://github.com/redblobgames/mapgen2/ However the html5 UI isn&
164.
by amitp 9y ago
Glad you liked it! It's still going, with new owners :) I'm revisiting the map generation because the authors of RotMG are working on a new game, and have asked me to work on the maps.
165.
by amitp 9y ago
All cool techniques — also see this paper about procedural terrain generation with GANs https://hal.archives-ouvertes.fr/hal-01583706v2
166.
by amitp 9y ago
If you're on a Mac, try Coconut Battery to check the battery health. Mine was showing 30% of original capacity, which explained why it kept draining quickly. Resetting (recalibrating?) it by letting it drain completely and letting it c
167.
by amitp 9y ago
I love hexagons too! Hexagon related but not related to the original article: - http://esolangs.org/wiki/Hexagony - a hexagonal layout programming language, implemented in Ruby, which is also the name a gemstone made o
168.
by amitp 9y ago
Yes, it works for me. I'm using the Mitsuharu version. Results: text mode - 6.7ms (SD=0.3); fundamental mode - 6.7ms (SD=0.2); js2 mode - 13.2ms (SD=11.6); org mode - 13.9ms (SD=14.2)
169.
by amitp 9y ago
You may find World Machine interesting: http://www.world-machine.com/about.php?page=features
170.
by amitp 9y ago
Meshes are used in various kinds of simulation calculations: https://en.wikipedia.org/wiki/Finite_element_method ; there's also discrete exterior calculus: http://brickisland.net/cs177fa12/?ca
171.
by amitp 9y ago
(I can't get to the site right now) They have a cool implementation that uses es6 generators to interrupt the algorithms as they run, and pass the algorithm state up to the visualization system. Other algorithm / data structure vi
172.
by amitp 9y ago
For me it was the other way around. The expensive school computer was locked down (this was the mid-1980s) but I learned stuff there that I then explored at home on a home computer. People tried to get into the school computer to capture ot
173.
by amitp 9y ago
I agree, reverse engineering some of this stuff is hard. I had been studying economics & systems design and put lots of interacting feedback loops into SRE so it's sometimes hard to tease out the individual effects. I've put s
174.
by amitp 10y ago
Title's misleading. A* is also 100x faster than A* ;) It depends a lot on the graph you give it. JPS by itself (not JPS+ in this video) is notable for not requiring any precomputation to achieve some speedup. If you're willing to
175.
by amitp 10y ago
(Author here) I had been trying to avoid the irregularities for that particular project because the storage of extremely large game maps saves a lot of space if you can use a regular grid instead of an irregular graph. (Certain algorithms l
176.
by amitp 10y ago
(Author here) I had been trying to hide the 12 pentagons on the game map but RimWorld stretches the hexagons so that it doesn't have to hide anything. You can see one here: http://steamcommunity.com/sharedfiles/fil
177.
by amitp 10y ago
HEALpix is cool. It has the same connectivity issue as with all the solutions — some locations are a little different. With hexagons a few polygons are different (polygons); with HEALpix a few vertices are different (three quads instead of
178.
by amitp 10y ago
(Author here) S2 is great (and Eric is brilliant). It uses a cube projection underneath. S2 solves some problems I didn't have (equal area, curve filling) but the main problem I was looking at was the nonuniform connectivity, and as fa
179.
by amitp 11y ago
Two of the many things I love about setosa: 1. The visualizations are designed specifically for each topic, instead of a generic visualization used for all topics. 2. There's a narrative explanation supported by visualizations, instead
180.
by amitp 11y ago
And if your maps don't change, you can use precomputation with A* , making it even faster than JPS+. 1. Preprocess the graph to make it smaller. Visibility graphs are a first step, but if you have grid movement you can remove the redun

More ›