New general-purpose optimization algorithm promises order-of-magnitude speedups

Very impressive work, here&#x27;s the arxiv link: <a href="http:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;1508.04874" rel="nofollow">http:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;1508.04874</a>

In an interesting coincidence, I have another of their papers in my hand, and it&#x27;s equally as impressive (I&#x27;m using it to improve a related solver in econometrics).<p>One problem though is that the constant terms of these models make take a while to go down (i.e. don&#x27;t expect to see this in production anywhere for at least 5yrs), but it&#x27;s still great to see progress there.

A very cool paper, esp how they are able to improve other algorithms. Each of those could have been a paper by themselves.

Wow, that is great news. solving optimisation problems is one of the true benefits of IT that few people know about. From warehousing to routing we use it every day in logistics and supply chain.<p>It is gradually moving out in finance and accounting from the big guys.

Are there implementations available, yet?

Did anyone else think it was strange, in an article about runtime efficiency, for them to say that problem-specific optimisations promise speedups of &quot;several orders of magnitude&quot;? Because, well, that doesn&#x27;t change the order of the algorithm...