First, we focus on the application of IAPG in risk scenarios, where decisions made by AI can, for example, endanger the physical integrity of people. Real examples will be taken in industry with critical facilities, where the idea is to learn with little data, or there are diverse tasks to be solved. The idea is to investigate the robustness of current IAPG models such as Few-Shot learning and AutoML, among others, and propose solutions to make them more reliable in this field:
Objective 1
Establish prototype scenarios
Objective 2
Assessing the robustness of IAPG methods
Objective 3
New solutions to improve the reliability of IAPG.
