Dynamic memory allocation for CMAC using binary search trees
Peter Scarfe, Euan D. Lindsay · eSpace (Curtin University) · 2007
Cerebellar Model Articulation Controllers (CMACs) are a biologically-inspired neural network system suitable for trajectory control. Traditionally, CMACs have been implemented using hash-coding for their memory allocation, requiring static allocation of fixed amounts of memory in advance to the training of the system. This paper presents a method for implementing CMACs using Binary Search Trees to provide dynamic memory allocation, allowing for lower memory usage without compromising the functionality of theCMAC.