AutoLearn-SI - Leveraging Benchmarking Data for Automated Machine Learning and Optimization.

Project Overview

The main goal of AutoLearn-SI project is to establish a permanent and excellent ERA Chair research group specializing in the AutoAI domain, especially in AutoML and AutoOPT, which will be pivotal in Slovenia and its neighboring region and will ensure sustained excellence, heightened visibility, and improved integration within the European Research Area. This strategic move not only enhances future competitiveness in securing EU research funding but also actively contributes to the promotion of national institutional reforms by integrating recent AutoML and AutoOPT advances into educational training. The novel ERA Chair will focus on leveraging benchmarking data for AutoML and AutoOPT. This will enable the fusion of knowledge and ideas for target-oriented research on three vertical pillars (Experimental databases, Representation Learning, and Automated Algorithm Selection/Configuration) and two Horizontal themes (Single-objective Optimisation and Multi-label Classification).

Experimental Databases

Representation Learning

Algorithm Selection / Configuration

Single-objective Optimization

Multi-label Classification

The AutoLearn-SI will be initially placed in the Computer Systems Department, with its permanent staff employed under the ERA Chair scheme, collaborating with staff from participating departments, the Computer Systems Department and the Department of Knowledge Technologies.

Objectives

1. To establish a leading AutoAI (AutoOPT and AutoML) research and educational center regionally through a new ERA Chair at Jožef Stefan Institute.
2. To educate and train young researchers, students, and young entrepreneurs through data-driven analyses work in approaches that automate the selection of methods and their hyper-parameters for two learning tasks, namely single-objective optimization and multi-label classification.
3. To provide training for young researchers, national students, and young entrepreneurs in transferable skills such as Science Communication, AI and Data Privacy & Security, and Patent Application, alongside the organization of stakeholder workshops.
4. To develop and validate open-access sustainable education and training material for AutoML and AutoOpt techniques aligned with the EU AI Act.
5. To strengthen the sustainability of the ERA Chair through scientific research collaboration on AutoLearn-SI knowledge with international organizations and engage with national stakeholders.

Implementation phases

The project will be implemented across three phases:

Initiation and Establishment Phase

(M1 - M15)

Acquiring and Extending Knowledge

(M16 - M36)

Sustainability and International Visibility

(M37 - M60)

Each phase includes two work packages (WPs): one concentrating on project management and coordination, networking, outreach, and sustainability activities (WP1,3,5), and the other emphasizing training through data-driven work-oriented and education activities (WP2,4,6).

Project Information