VERFAHREN UND VORRICHTUNG ZUR BESTIMMUNG DER GLOBALEN SPEICHERGRÖSSE EINER GLOBALEN SPEICHERZONE, DIE DEN DATEN EINES NEURONALEN NETZES ZUGEORDNET IST

摘要

Method for determining the global memory size (MaxF) of a global memory area (MF) to be allocated in a memory intended to store input data (M1) and output data (M2) of each layer (L) of a neural network (RN), the method comprising for each layer (L) a determination of an elementary memory size of an elementary memory area (ZA) intended to store input data (M1) and data output (M2) of said layer (L), said elementary size being between the memory size of the input (M1) or output (M2) data of said layer (L) and a size equal to the sum of the memory size of the input data (M1) and of the memory size of the output data (M2) of said layer (L), the overall memory size (MaxF) being determined from the elementary memory sizes (ZA) associated with said layers (L), the global memory zone (MF) containing all the elementary memory zones (ZA).