You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thanks for the feedback! I'm so glad to have people taking an interest!
At one point, I had a script for benchmarking some core operations (add, multiply, solve Ax=b, find eigenvectors) for matrices and vectors of different sizes, but there were no comparisons to similar libraries. Eventually I dropped it because it wasn't especially useful in its current form, but that code could still be a good starting point for adding in benchmarking that is actually useful.
Especially on the performance improvement front, we could also do things like operation counting to verify that there are no extraneous operations - it would just take a new subclass of ScalarOperations.
There are probably lots of opportunities for performance improvement - so far my focus has just been on having fun implementing different decompositions and ML algorithms, but some time in the near future I definitely want to take a stab at making this at least as fast as other comparable js libraries.
Introducing benchmarks against other linear algebra libraries could:
An example of such library could be: https://github.com/sloisel/numeric/ it's dead, but still widely used today.
Thank Your for your awesome project!
The text was updated successfully, but these errors were encountered: