What is the future trajectory of HL?

[Bern]

Hello HL folks,

I am writing a paper comparing various machine learning techniques and genetic algs and symbolic regression in particular, and hope to feature HL. However I am concerned about recommending packages that are fading away. (As we all know, the internet is full of abandoned projects that still have a beautiful download page. DEAP, for example, seems mostly dead).

Looking at HL it seems that there hasn't been a new release in more than 2 years. However I also notice regular updates on their changelog (https://dev.heuristiclab.com/trac.fcgi/timeline). And this google group seems pretty active.

How would you characterize HL? A vibrant project with growing momentum? A project that is treading water? A project on its last legs of residual support, where the developers have since moved on?

Regards,

Bernard

[HeuristicLab]

Dear Bernard,

 

thank you for your message and your interest in HeuristicLab. I think we can best characterize HeuristicLab as a project which reached a phase of stability. We – the people who are mainly responsible for HeuristicLab – are the research group “Heuristic and Evolutionary Algorithms Laboratory” (HEAL, https://heal.heuristiclab.com) located at the Hagenberg Campus of the University of Applied Sciences Upper Austria. With about 20 members we are one of the largest research groups of our university and many of us use HeuristicLab frequently in our projects and lectures. Furthermore, as far as I know, HeuristicLab is also used at several other universities in research and education around the world, although I do not have any recent numbers. Therefore, HeuristicLab has a very active user base, which is also documented by the activities on our mailing list.

 

Nevertheless, as you mentioned our last release was more than two years ago and it is true that the development speed of HeuristicLab slowed down in the last years. The main reason from my point of view is that HeuristicLab offers most of the features that we need for our daily work. We spent many years on making HeuristicLab a piece of software which we are proud of and which serves us well in research and teaching. Due to the great effort of many people – and I deeply thank everyone who contributed to HeuristicLab – I think we have been very successful in this regard, as HeuristicLab has become a very powerful application for heuristic optimization and data analysis.

 

However, please do not get me wrong. Although HeuristicLab reached a high level of maturity and provides most of the features that we regularly need, it does not mean that its development has stopped. Many of the developers who created HeuristicLab in the last 15 years are still members of our group and we also continue our work on it. For example, we will provide a new release within this year and we are also currently working on new concepts for symbolic regression and for dynamic optimization, as you can see in the timeline on the website. Furthermore, the next release will also contain a new JSON-based command line interface which will make it much easier to use HeuristicLab without its GUI and to integrate it with other applications. Additionally, we are also working on fundamental topics such as migrating to .NET 5, pushing the development process to GitHub, and implementing a completely new version of HeuristicLab Hive. So stay tuned, there is more to come …

 

All the best and kind regards,

Stefan

[Bern]

Thanks so much for this detailed reply Stefan, this answers my question very well. (That someone even replies to this message is already 50% of an answer!)

And of course I wish the whole endeavor the best of luck!

I also welcome the addition of the JSON interface - I assume that this means it will be more interface-able from e.g. Python - as this is one of the shortcomings of HL that is currently impeding its use in my paper.

Regards,

Bernard

On Friday, September 24, 2021 at 10:08:52 AM UTC-4 HeuristicLab wrote:

Dear Bernard,