Artificial Neuron Networks (NNs) are a main driver of applications of Artificial Intelligence. However, their ubiquitous deployment is hindered by their large requirement of computation resources, caused by the vast amount of operations required to execute them and the large amount of memory that requires to store the model and intermediate data. This work optimizes NN applications by tackling this problem at three levels of abstraction: the algorithmic level (what needs to be calculated), the dataflow level (in which order are things calculated) and the hardware level (which circuits calculate it).
Defence starts at 17.00h.
26/4/2023 17:00 - 19:00
Aula Jozef Heuts, Landbouwinstituut Hoofdgebouw, Kasteelpark Arenberg 20, Heverlee
