Algorithms and Software Technology (AST)
Algorithms and Software Technology is one of the three research programme of the Leiden Institute of Advanced Computer Systems (LIACS). We perform fundamental research in the areas of algorithm design and analysis. Our emphasis is on algorithms and architectures for mining large data volumes as well as on natural computing. In these areas, we focus on the development of formalisms, methods, techniques and tools to design, analyse, and construct software systems and components.
The research programme contains two themes.
Computer science has its roots in mathematics. That is why LIACS' theoretical research is undertaken by a team of computer scientists and mathematicians. Our efforts are focused on better understanding the fundamental characteristics of specific computational problems.
New and correct algorithms
Our work regularly results in new, faster algorithms. Often, due to their wide range of applications, these new algorithms lead to across the board improvements in for example health care, industry, and biotechnology. To further ensure safety and dependability in mission-critical applications, like the auto pilot of an airplane and the controller of a power plant, we also study the correctness of software and algorithms. By creating algorithms that verify other algorithms (automating prove techniques for software correctness) we exclude harmful bugs that may never be revealed by testing alone.
One subject that enjoys much attention at LIACS is concurrency. In modern information systems, a large number of different components are often active simultaneously, leading to a huge variety of complicated interactions. The aim of our research into concurrency is to understand and model these interactions and predict their effect on the global state of the system. In collaboration with biologists for example, we produce models of the many processes that take place in live cells. As a result, it becomes possible to predict how a cell will respond to the introduction of an external agent, such as a drug, without doing an in vitro experiment.
We are also involved in research into coinduction. Coinduction is a relatively new mathematical technique that deals with circular mathematical structures. Due to their circularity, these structures can capture infinite systems, comparable to the ‘Droste effect’. For this reason, coinduction can be used to abstract various complicated real-time and probabilistic systems, in order to prove properties over all of them in one strike.
Computers are capable of formulating new algorithms on the basis of data they themselves have gathered. In other words, these computers can learn without having been pre-programmed by humans. They make predictions we never expected. At LIACS, we explore the possibilities offered by this revolutionary new generation of computers.
Data contain information that is hard to extract for humans, but can be sieved out by computers. Knowledge Discovery from Data (KDD) concerns the identification of patterns that capture the structure underlying the data. Subgroup discovery, for example, is a topic in pattern mining investigated at LIACS that has been successfully applied to fraud detection and sports analytics.
Every production process is in essence a search for the best combination of a variety of factors, each with their own individual optimum. The art lies in identifying the overall optimum for all of those factors combined. In 2016, our researchers developed a new method for the theory underlying this problem, known as ‘multi-objective optimisation’.