150 points by opti_queen 6 months ago flag hide 7 comments
efficientcoding 6 months ago next
This is really impressive work. I've been struggling with large scale optimization problems for ages. Can't wait to try this out!
optimizeorbust 6 months ago prev next
Just read the research paper. This is certainly a game changer! It's a pity that the code isn't open-sourced yet, though.
sr_coder 6 months ago prev next
The author puts this approach in the realm of two existing optimization techniques, but I'm a bit fuzzy on the implementation details. Can anyone clarify?
dnn_enthusiast 6 months ago next
Yes, it seems like there's a hybridization of Stochastic Gradient Descent and the Dual-Simplex method. This should be interesting to explore.
mscodeburn 6 months ago prev next
I think the real advantage is the new initial point selection process. Has anyone measured its effect on different datasets?
deeplearner 6 months ago next
According to the paper, this method can provide a 30% performance improvement on properly scaled datasets. But I'm curious about its efficiency with small to medium scale challenges.
advancedalgos 6 months ago prev next
Anyone working on adapting this for quantum computing applications? It'd be fascinating to see how this performs with QUBOs and Ising models.