More manpower for Finland's AI research - Technology Center Program NVAITC

More manpower for Finland's AI research - Technology Center Program NVAITC

NVIDIA, the Finnish Center of Artificial Intelligence (FCAI) and CSC have just published their collaborative technology center program (NVAITC). NVAITC is an abbreviation for the “NVIDIA AI Technology Center”.

The aim is to promote Finnish AI based research expertise by increasing expert resources and ensuring sufficient GPU-based computing capacity for research projects and access to the best expertise related to the use of GPU technology in machine learning.

Resources assigned to cooperation from all parties

Each party is committed to providing human resources to implement the NVAITC program; for example, the Finnish NVIDIA has hired a full-time employee for this. Through FCAI, the parties involved in the cooperation include VTT Technical Research Centre of Finland, the University of Helsinki and Aalto University as well as their research groups. The co-operative program also involves AI based training as a part of NVIDIA's Deep Learning Institute (DLI) program.

The origins of the technology center program

The NVAITC program led by NVIDIA originates from Singapore where it has strengthened AI research between companies and higher education institutions. So far in Europe, the program has been launched in Luxembourg and Italy; as the third country to join, Finland is among the first ones.

The program helps identify promising AI research projects in order to involve top international technology experts at NVIDIA to support the research groups. The NVAITC center is particularly needed for research projects that require a great deal of computing capacity and machine learning expertise. These may include artificial intelligence applications related to medicine or smart cities. Our Puhti-AI platform, based on NVIDIA's GPU technology, is used as a platform for GPU calculations. The requirements for project approval include creating a scientific publication (with a mention of NVIDIA).

First projects already in progress

One of the first projects is the development of methods based on differential privacy, conducted by Antti Honkela’s (HY) group. In addition to pure anonymization, these methods can be used to carry out analyses whose results can be guaranteed not to violate anyone’s privacy. NVAITC contributes by profiling the code performance model for different topologies and developing a strategy to eliminate bottlenecks.

Professor Jaakko Lehtinen's (Aalto) group is studying GAN networks (Generative Adversarial Network). The study includes feature optimization calculations such as dimension reduction and density assessment with a large number of training examples. GAN is a key family of generative modeling algorithms; for example, the GAN network can be used to create synthetic images of "invented" people (such as celebrities.) The project presents significant computational requirements, and together with NVAITC, it aims to improve the ability of multi-node training and thus enable the introduction of more extensive Puhti-AI resources by CSC.

This is an important partnership program for CSC, bringing our extensive PUHTI AI computing capacity into use by FCAI-led research groups. FCAI is a significant actor in the Finnish field of research of machine learning, as is demonstrated by its flagship status of the Academy of Finland, a status which secures significant financial support. One objective is to intensify cooperation, thereby promoting the implementation of the Finnish RDI map

More about this topic » Go to insights and news »

Pekka Uusitalo

Writer is responsible for developing partnerships at CSC.

pekka.uusitalo@csc.fi