Nowadays, machine learning is very present in the public debate. Central Processing Units (CPUs) or Graphics Processing Units (GPUs) can perform the computations for machine learning algorithms but are not very efficient and consume a lot of energy. Therefore, we need to create dedicated hardware that is built specifically for neural networks. This dedicated hardware should be small enough (to keep it cheap), efficient for fast execution (for real-time applications) and low energy (to avoid battery problems). This work focuses on the hardware architecture search for neural networks, hardware-algorithm mapping and tools to optimize this.
10/1/2024 17:00 - 19:00
Aula Jozef Heuts (LAND 00.215)
