Daniele Ielmini, professor at the Department of Electronics, Information and Bioengineering, will conduct ANIMATE (ANalogue In-Memory computing with Advanced device Technology), a project that aims to develop a new computational concept to reduce energy consumption in machine learning.
We generate, process and use a huge amount of data every day. Searching for a keyword on the internet, choosing a film for the weekend or booking our next holiday are just some of the actions that rely on data-intensive algorithms.
The energy cost of this type of calculation is extremely high: it has been estimated that training a conventional neural network for artificial intelligence (AI) produces the same amount of carbon dioxide as 5 cars in their life cycle. Data centres, which currently meet most of the world’s AI needs, now consume about 1% of global energy demand, with growth expected to reach 7% by 2030. To correct this worrying trend, new energy-efficient hardware solutions are needed. Professor Ielmini’s preliminary ANIMATE research has shown that computational energy requirements can be reduced by closed-loop in-memory computing (CL-IMC), which can solve linear algebra problems in a single computational step.
In CL-IMC, the time to solve a given problem does not increase in proportion to the size of the problem, unlike other computing concepts, such as digital and quantum computers. Thanks to the reduction in calculation time, CL-IMC requires 5,000 times less energy than digital computers with the same accuracy in terms of number of bits.
Ielmini’s project will develop the device and circuit technology, system architectures and set of applications to fully validate the CL-IMC concept. System-level architecture and exploring its applications will further prove the scalability and feasibility of the concept, to prove that CL-IMC is a major contender among energy-efficient computing technologies.
Our university once again proves to be at the forefront, having outperformed its scholarly competitors in a very competitive selection process, with only 14.6% of the 1735 projects submitted receiving funding. With this project, the Politecnico di Milano has been awarded a total of 86 European Individual Grants (including ERC and Marie Curie).
