It’s a good question. We need to improve our (box) constraint support for example. I doubt we do BFGS faster as well. However, our Julia implementations have several advantages over Matlab.

– It’s easy to switch between line searches for example, and you can easily add your own.

– The code is publicly available (I’m not sure if some of the Matlab code can be checked?), so you can always have a look, improve things, and check for bugs. If you find one, it’s easy to file and issue or even fix it yourself.

– Say your optimization problem is related to some image processing, then you may want to represent numbers as Float32’s, and save memory that way. That is super straight forward in Julia. You can also add your own “real” numbers and optimize over them, as long as you implement the necessary methods.

– …

In many ways I expect Matlab to have worked long and hard enough on their implementations such that is it hard for us to “beat” them, but we provide the usual flexibility and ease of extensions that Julia often brings to the table.

We’ll do some proper comparisons in the future.

]]>How do you compare it to MATLAB’s Optimization Toolbox feature wise and speed wise?

Thank You.

]]>I will make sure to provide some more details next time.

]]>Thank you so much for sharing your code with Julia community. I appreciate your contribution as we dont have a good number of resources to go forward with Julia. I am following your courses. May I request you to share/post all the mathematical(equations) solutions along with Julia codes. It will help up to follow your courses from the USA (or from any corner of the World). I would request you to share all of your course materials(if your institutional policy permits) to those who are interested. Thanks ]]>