OPTANO Algorithm Tuner

OPTANO Algorithm Tuner Version 1.0.0 – for even faster and more efficient algorithms

 

The OPTANO Algorithm Tuner is a highly sophisticated tool which uses an evolutionary machine learning process and mathematical optimization to enable you to configure an existing algorithm so that it is tuned precisely to the corresponding use case or a particular type of input data you already have. To put this more simply, we’d like to use an analogy that our developers used in an interview in early summer2020: As you know, in Formula 1 racing, a great deal of tuning is done on the cars, depending on how straight the racing track is, how many tight curves it has, or what the weather conditions are like, etc. That’s why the car is retuned and tested before every race.

So your algorithm is the car and we can tune the parameters in different ways, depending on what the algorithm is going to be used for. In other words, the OPTANO Algorithm Tuner lets a lot of Formula 1 cars be driven and it looks to see which combination of parameters is best for the problem at hand. While algorithms can have very long run-times in general, the OPTANO Algorithm Tuner retunes your machine learning and optimization algorithms again and again until they run as fast and efficiently as possible.

Der OPTANO Algorithm Tuner Version 1.0.0. – was ist neu?

When we first released our OPTANO Algorithm Tuner back in summer 2020, we were certain of two things – first of all, that we had released something really good and secondly, that we could make it even better. And it seemed that the more we worked on the Tuner, the more ideas we had. Over the last few months, we’ve been putting these ideas into practice. We have completely rebuilt the inner architecture of the tuner to make it more elegant and leaner – the result being the current OPTANO Algorithm Tuner 1.0.0 with two essential improvements: even faster performance and improved customizability.

OPTANO Algorithm Tuner

Parallelization of the tuner has been optimized so that Machine Learning and optimization algorithms can now be tuned even more precisely and, above all, even faster. What’s more, improved customizability in Version 1.0.0. offers even more possibilities to optimize tuning for an individual tuning target. Version 1.0.0 is now up t five times faster than the previous version 0.9.1. (see diagram opposite).

In a nutshell: Thanks to optimized parallelization and customizability, the new Tuner version is now up to five times faster than its predecessor. This is evident in the graph below:

OPTANO Algorithm Tuner - 0.9.1 vs 1.0.0

Usually the tuner is run for a certain number of generations (mostly 100). The plot shows the time required over the number of finished generations. Thanks to its improved parallelization and customizability, in this dataset, Version 1.0.0. is five times faster than Version 0.9.1.

“Do you want to test the OPTANO Algorithm Tuner?
 

What are the user benefits?

Algorithms are usually a time-consuming business and usually need several weeks until they are configured. This can be detrimental to projects or businesses where the time factor plays a critical role. Even faster tuning as well as better customizability offer the following user benefits:

Higher efficiency

No matter what you are planning, whether it’s delivery schedules, production processes, job scheduling, etc., when your algorithms run faster, your business can find better solutions within a shorter space of time thus saving you a considerable amount of time and resources

OPTANO Algorithm Tuner
OPTANO Algorithm Tuner

Customizability:

The OPTANO Algorithm Tuner is highly adaptable and can be customized to meet your individual use cases or requirements.

Distributed Execution:

The OPTANO Algorithm Tuner is highly scalable to your IT-infrastructure. You can distribute its computations on several nodes in your cluster and speed-up its execution.

OPTANO Algorithm Tuner

Last but not least…

Sometimes it is actually better to leave certain things unchanged. Which is why Version 1.0, just like the previous version, is  still open-source and now available for download free-of-charge at NuGet or GitHub. Why not download it here and give it a try?