OPTANO Algorithm Tuner

OPTANO Algorithm Tuner Version 2.0. – now with gray box optimization

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. This is a valuable asset to businesses since more efficient algorithms lead to faster and optimal business solutions.  We recently explained in detail how the Tuner works  in a recent article which you can read here.

Even though we are particularly proud of what we have achieved with the OPTANO Algorithm Tuner so far, we haven’t been resting on our laurels. Behind the scenes, our developers have constantly been researching, developing, testing and refining in order to get the very best out of the Tuner’s potential and we believe that with the release of 2.0, we are closer to our goal than ever before….

The Gray Box: an innovative idea

As you may already be aware, configuring algorithms is time-consuming and can take days, sometimes even weeks. A tuner which considerably speeds up this process is a huge business advantage and the Algorithm Tuner has performed even faster and more efficiently with each new version we have released. To evaluate target algorithms, we have previously used an approach in which a target algorithm evaluation is started and which makes use of its final result only after the evaluation has been completed. (black box approach). In our current version 2.0, we can now offer an innovative alternative to the standard black box: gray box optimization.

Gray is the new black

We already mentioned above that the black box approach only allows you to consider the final result of a target algorithm run. Hence, the evaluation of the target algorithm will still keep running right up until completion time, even though it may be a sub-optimal configuration.

By applying the gray box, however, you can detect sub-optimal configurations via intermediate target results at runtime. In order to do this, you can set the intervals to whatever you wish – be they 5 minutes or 30 minutes. As soon as the gray box detects that a configuration is not optimal for the required use case, it will automatically stop the runtime. We can compare this with Darwin’s “survival of the fittest” theory. As with Darwin only the most optimal configurations will survive – or cross the finish line, so to speak –  as the OPTANO Algorithm Tuner only focuses on promising configurations.

The diagram below provides an overview of how the gray box approach works in comparison to that of the black box:

Gray Box vs Black Box overview

 Overview of the Black Box vs Gray Box

 Do you want to try out the OPTANO Algorithm Tuner with the gray box?

The results of the gray box tuning can be seen in the following diagram quite clearly: While both approaches surpass the default configuration and converge to the same “fitness”, so to speak, the gray box approach remarkably speeds up the complete tuning system as it can find good configurations faster than its black box counterpart. Tests have proven that with the gray box, Version 2.0  runs 1.5 faster than its predecessor.

Gray Box vs Black Box in test

The gray box: when time is everything

Since the gray box considerably speeds up the solution process of OR solvers this offers a major advantage when it comes to solving optimization problems where the time factor is critical. Furthermore, it considerably improves the solution quality of the heuristics (problem-solving techniques) which are applied.

Take for example the following scenario:

A company offers facility management services within a 200 km radius. It employs about 250 service engineers. Clients will phone the call center to make appointments for the engineers to come out and make the necessary urgent repairs. The company’s goal is to streamline and speed up the scheduling process. It wants to ensure that emergency appointments take place within the shortest time possible after the call center has been notified of the problem, so the shortest route to the client needs to be chosen.  This increases the productive time at the client’s premises, thus guaranteeing increased customer satisfaction. On the other hand, by quickly being able to choose the optimal route every time, journey times and running costs can be minimized in the long term.

This is where the Tuner can be of enormous help. By tuning the algorithms at the company to make them even faster and more efficient, it can find better solutions within a much shorter space of time thus saving the company a considerable amount of time and resources.

And this not only applies to scheduling problems. The OPTANO Algorithm Tuner can also be of advantage in many other areas where speed and efficiency is critical: production planning, network planning, and many more.

By the way…

Just like our previous versions, Version 2.0 with its gray box optimization is just as adaptable as before, meaning it can  be customized to meet your individual use case requirements. Finally, it is still open-source and available for download as of now, free-of-charge at NuGet or GitHub Why not try it out today?