I was using this as a reference as I was working on a Rust implementation accessible in JS via FFI. However when comparing values it appears this implementation doesn't actually follow the paper and computes a sub-optimal assignment.
I randomly generated a 5000x5000 cost matrix, exported it to JSON, and then timed each implementation against solving the same matrix.
Pure Rust:
Calculated in 3.64231825s
Total Cost: 11,216.864157676053
Rust via super inefficient NodeJS FFI :
Calculated in 10.239015541911126s
Total Cost: 11,216.864157676053
This Script:
Calculated in 612.758542060852ms
Total cost: 30,898.483242653267
The cost is almost 3x as high and the runtime is almost 3 seconds faster than a native compiled language. I haven't dug super deep into the code, but it appears that this isn't a correct implementation of the algorithm.
Fil
commented
1 year ago
I was using this as a reference as I was working on a Rust implementation accessible in JS via FFI. However when comparing values it appears this implementation doesn't actually follow the paper and computes a sub-optimal assignment.
I randomly generated a 5000x5000 cost matrix, exported it to JSON, and then timed each implementation against solving the same matrix.
Pure Rust: Calculated in 3.64231825s Total Cost: 11,216.864157676053
Rust via super inefficient NodeJS FFI : Calculated in 10.239015541911126s Total Cost: 11,216.864157676053
This Script: Calculated in 612.758542060852ms Total cost: 30,898.483242653267
The cost is almost 3x as high and the runtime is almost 3 seconds faster than a native compiled language. I haven't dug super deep into the code, but it appears that this isn't a correct implementation of the algorithm.