I.3.7. Establishment of the Centre for Machine Learning (ML)

IDUB Programme Objective(s)

• Increasing the impact of scientific activity conducted at the University on the development of world science.
• Improving the quality of education for students and doctoral candidates in fields related to the priority research areas (POB).

Name of POB/activity group

• II. Beyond Micro and Macro Worlds

Description

Objective

The objective is to increase the research potential of UW and improve researchers’ efficiency in the use of machine learning methods, by creating a new group of data scientists to provide technical assistance in constructing machine learning models. Providing ready-to-use models will significantly reduce the burden of mastering advanced computer techniques and data processing, which now affects those planning to conduct research using machine learning.

Justification

Machine Learning (ML) allows for more effective data analysis in a wide variety of research areas, including many of those conducted at the University of Warsaw. The main obstacle preventing research teams from incorporating ML in their current/planned activities is the need to master advanced computer techniques in order to complete the necessary “training” process. This requires employing a specialist or designating a team member to learn the procedures. In the latter case, the individual is forced to deal with issues which are unrelated to the central question being studied. Many technical aspects related to ML are universal (defining a data stream, efficient calculation, efficient use of computing resources, the ability to use advanced computing techniques in the cloud). Another issue is access to computing infrastructure. Training models for ML require considerable computing power and may demand non-standard devices, e.g., graphics cards (GPUs). Purchasing such equipment is not an optimal solution: a much more efficient approach is to use the existing infrastructure offered by specialised units, or to purchase computing time offered by commercial suppliers. Yet each of these options requires appropriate knowledge and experience.

Methodology of operation

The Centre’s staff will follow the ongoing developments in the area of ML and transfer its know-how to research teams through cooperation; this will consist of preparing and launching ML tools for a given team. Whenever possible, the Centre’s staff will create a working code to process data with ML and obtain the preliminary results for the researchers. The selection of projects to be worked on by the Centre will be based on a continuous call for proposals. In the first stage of the Centre’s operation (in 2020), the implementation of four applications is expected – one from each department involved.

In addition to cooperating with specific research groups, the Centre will organise training sessions in how to use ML tools.

Tasks

1. To gather knowledge and experience in the use of ML-related tools:
   1. monitoring progress in ML through participation in ML training and workshops;
   2. increasing the competence of the Centre’s members through the application of standard ML elements (e.g., input-output streams, the division of calculations, and data representation) in various techniques;
   3. becoming familiar with the various opportunities for accessing the computing infrastructure, in particular those offered by the PL-Grid project, as well as commercial cloud computing services such as Google Cloud or Amazon Web Services;
2. To transfer know-how to research groups and all other interested persons at UW:
   1. organising workshops on ML for UW once or twice a year;
   2. participating in the work of research groups by directly involving or supervising students during the use of ML;
   3. increasing the visibility and importance of existing initiatives which are related to ML (funding for the PLinML conference), which has been organised for several years at MIMUW; organization of a lecture on ML for a wide audience at PLinML, modelled on the “Ask a Physicist” lectures which are held at the University of Warsaw.

Coordinating unit

• Faculty of Physics

Entities involved in implementation

• Faculty of Physics
• Faculty of Chemistry
• Faculty of Biology
• Faculty of Mathematics, Informatics and Mechanics