Mappost collaborates with IMCS UL to help link theory and practice of combinatorial optimization

Mappost continues its collaboration with the University of Latvia Institute of Mathematics and Computer Science (IMCS UL).

One of IMCS UL’s research areas is the applications of neural networks in combinatorial optimization, and together we're looking forward to linking this with the more practical direction of combinatorial optimization using tools like OptaPlanner.

Lately, we've been discussing practical problems and challenges related to business requirements and technical issues. Our goal is to explore practical niches where IMCS UL scientists can apply their methods and technologies related to neural networks. By bridging the gap between theory and practice, we hope to build things in the right direction. 

In the past, we have collaborated with IMCS UL to solve particular challenges. Recently we held another session where we shared challenges and results from our practical experience while enriching each other's knowledge. 

Stay tuned for more updates on this exciting collaboration! 

About Mappost: a route and resource optimization software development company working with companies in industries like waste management, vending, postal services, forestry, agriculture, and others.

About IMCS UL: the institute was established in 1959 and currently consists of more than 200 employees from whom about half is in academic positions. Annualy publishes approximately 122 scientific publications and apart from research, IMCS UL has always been involved in commercial software development.

About project ”Combinatorial optimization with deep neuron networks” (No.lzp-2021/1-0479): The project aims to develop better combinatorial optimization algorithms using deep machine learning methods. It will solve complex problems and explore whether neuron network-based approaches can outperform hand-crafted solvers. Deep learning models suitable for combinatorial optimization tasks will be developed, and their capabilities will be evaluated on both synthetic examples and real-world cases, in collaboration with partner companies. Results will be presented at high-level AI conferences, including three new technology prototypes and code on GitHub. More about the project here.