EliteForsk grant for young ITU researcher in search of the optimal search algorithm

PhD student Tobias Christiani has received an EliteForsk travel grant of DKK 200,000. His next stop will be Harvard and MIT in search of the holy grail of algorithms: the optimal search algorithm - if it exists.

Computer ScienceResearchgrants

Tobias Christiani is one of the promising young researchers who will receive a travel grant of DKK 200,000 from EliteForsk on Thursday, February 23. He is currently writing his PhD on similarity search, one of the most well researched and notoriously difficult problems in computer science. The problem is about finding efficient search algorithms that find similar objects in a collection of images, sound files or text documents.

The grant from EliteForsk will fund a research stay in Boston, where Tobias Christiani will work with some of the leading researchers within the field at Harvard and MIT this fall.

The holy grail of algorithms
Existing algorithms for similarity search are often both slow and bulky when applied to large datasets. Tobias Christiani is working to develop new and faster search algorithms and ultimately hopes to find what many consider the holy grail of the field: an optimal search algorithm that is effective regardless of the type and amount of data it is applied to.

»

My goal is to find a simple algorithm that always adapts optimally for what it searches in, and is so simple that a programmer will be able to implement it in half an hour.

Tobias Christiani, PhD Student
«
"There are theoretical algorithms that can reduce time and space consumption by adapting themselves to the dataset, but unfortunately they are quite complicated to implement in practice. My goal is to find a simple algorithm that always adapts optimally for what it searches in, and is so simple that a programmer will be able to implement it in half an hour. That would be huge. But we do not actually know if it exists," says Tobias Christiani.

Searching for theoretical proof
At the same time, he hopes to find theoretical proof that a more efficient algorithm does not exist.

"It is a bit like proving that a car can never go faster than 100 kilometers per hour. If that was the case, you could stop trying to build a faster car, shut down an entire research branch and focus on something else instead," says Tobias Christiani.

Although his research is within the theoretical corner of computer science and not directly linked to specific uses, several companies have shown an interest in the algorithms Tobias Christiani and his colleagues are developing in the Scalable Similarity Search research project. Their search functionality can be used for instance in internet search engines, in machine learning and for big data handling.

Further information

Vibeke Arildsen, Press Officer, phone 2555 0447, email viar@itu.dk