Thread 106236061

>Chinese scientists have developed the best shortest-path algorithm in 41 years! A team from Tsinghua University has broken Dijkstra’s “sorting barrier” — the first improvement since 1984.
>The new algorithm runs in O(m \log^{2/3} n) time. Potential applications? Faster shorter waits for route calculations, fewer traffic jams, cheaper deliveries, and more efficient computer networks. And, of course, a need to update computer science curricula
Based China!:
https://x.com/lukOlejnik/status/1955204273421128155?

>asian science
Probably fraudulent like everything else they publish

Dijkstra?

>>106236061 (OP)
Seems way too complex

>>106236061 (OP)
This looks like fucking math, where is the code?

>>106236061 (OP)
CHINA STOLE IT!!!

>>106236061 (OP)
>BMSSP
what a great name. it would be better if they name it after some chinese

>>106236318
Hey bro. Go look into this book. See how much "code" is inside it. I'm sure you'll find a lot.
(This book is "read" by a lot of university students btw)

>>106236348
this

gollygee we wuz inventorz n shiet

>>106236318
That is the code retard. It's pseudocode using mathematical notation (sets, etc). Most CS academic papers use notation like this.

>>106237529
>>106237742
Fuck that shit, gimme code, I want readable syntax not that abomination from the dark ages.

>>106236061 (OP)
I saw some guy say that it's only really faster on giant graphs

>>106237857
true they should use lisp notation.

>>106236061 (OP)
>fewer traffic jams
How would calculating a route 3ms faster stop traffic jams?

No those high brow academics should get down to earth if they pretend to love humanity so much, the world is imperative.

>>106237857
TSMT DESU
There's literally no reason for mathfags not to write python or at least something with actual descriptive variable names rather than single letter meme shit

>>106237857
Show some empathy. The average math fag is about as proficient at writing code as someone in their 3rd week of bootcamp.

>>106237742
Only that when it comes to actual improvements theoretical pseudocode is almost always bullshit because in actual code the overhead required for certain operations that are handwaved away with pseudocode actually push the performance so far down that the theoretical advantages disappear. The other anon is right, code with actual stats and comparisons or nothing, I've learned this the hard way trying to replicate many papers on GPU algorithms, in almost all cases good implementation of a "worse" algorithm significantly beats theoretical big O improvements.

>>106238041
you can port pseudocode to literally everything

Bigger than any LLM slop news ever if true

>>106236061 (OP)
Nothingburger