FCAI-originated open source software now gathered on the same webpage

The list, which will from now on be continuously updated, aims to bring top research results to broad use. Researchers and industry are the main target audience.

Open source software packages stemming from FCAI research have now been gathered on the same webpage.

Assistant Professor Arto Klami from the University of Helsinki is in charge of the page. He hopes that the list will help AI researchers find, and take into use, newest software tools as soon as they are published. The page also offers concrete tools for researchers in other fields, as well as for industry.

“We have collected a broad selection of FCAI-produced programs on the page, and especially highlighted ones that are easy to use for everyone”, Klami says.

As examples, he mentions Stan – which has already been taken up by at least a hundred thousand users, probably more – and ELFI, which is hoped to become a similar success story.

”Stan is the tool that most people today use to develop probabilistic AI methods. FCAI has played a significant part in the development of the core Stan software. Our list also includes many software packages for the Stan ecosystem that stem entirely from FCAI.”

Potential users for ELFI, on the other hand, come from other research fields as well as companies that use simulator-based methods in their operations. The tool is designed to make it easy to integrate data-driven elements for existing simulators.

Categorization makes differences between software tools visible

The software listed on the website have been divided into three groups. Highlighted software includes programs that represent the height of FCAI research, or programs that already have a broad user base.

“These are high-quality, easily usable, and central tools. Most researchers that visit the website already recognize some of these names”, Klami says.

The software listed in the General use category are of the same high level as the previous. They serve both companies and researchers from other fields.

The research software category, on the other hand, lists software that stem from ambitious projects but are so far not easy enough to use to serve all user groups.

"With this category, we are primarily reaching out to other AI researchers.”

Research findings to clear ecosystems

The software listing implements, in its part, FCAI’s goal to bring research results to as broad an audience as possible. Klami says that FCAI goes about fulfilling this mission with exceptional ambition.

"The principle is that an academic publication is accompanied by software; that is the standard in high-quality AI research today”, Klami explains.

”But FCAI goes beyond this traditional good practice by striving particularly hard to develop tools that are easy to use for a large audience.”

This emphasis can be seen in FCAI’s Highlight program A, whose sole aim it is to help bring research results from FCAI’s research programs to a broad user base.

"We do everything we can to make sure that all main results from the technical research programs are brought into as broad a use as possible”, says Klami who leads Highlight A.

To make this work as easy as possible, FCAI’s new research results are to a large part brought into clear ecosystems. In this way, new software do not remain scattered in research papers.

Klami says that, out of FCAI’s seven research programs, already three have easy-to-use environments into which most of the results are brought. For R1 this is the Stan ecosystem, for R2 ELFI, and for R4, Twinify and d3p.