Towards green AI‐based software systems: an architecture‐centric approach. GAISSA

Project

GAISSA is an interdisciplinary project. It includes aspects of various disciplines:

  • Artificial Intelligence, with focus on green AI;
  • Software Engineering, with focus on software architectures and software quality; 
  • Sustainability, with focus on energy efficiency and ecological transition on ICT. 

 

 Hypothesis

When modelling and developing green AI‐based systems,

the impact of architectural decisions on energy efficiency must be better understood, defined, reported, and managed

in order to deliver AI‐based systems with less demanding computational power needs

 

GAISSA Approach



AI engineering lifecycle consists of four stages also organized as cycles:

  • Data Management,
  • Modelling,
  • Development and
  • Operation.
GAISSA focuses on the two middle stages: Modelling and Development.

GAISSA Objectives:

  • Objective O1 sets up a greenability quality model for characterising and measuring the greenability (in terms of energy‐efficiency, resource optimization, ...) of both AI models and AI‐based systems.
  • Objectives O2 and O3 aim to provide green‐aware architecture‐centric methods to data scientists and software engineers, respectively.
  • Finally, objective O4 aims to support the analysis, decision‐making and reporting on greenability of AI models and AI‐based systems, so that data scientists and software engineers can use (i) advanced analytics functionalities and (ii) tangible assets.