Forum Discussion
Altera_Forum
Honored Contributor
14 years ago --- Quote Start --- My idea of the whole process consits of the following steps: 1. The new structure of the neural network is downloaded by the NIOS processor from the PC and implemented on the rest of the FPGA. 2. Weights of the neural network are modified by means of a Hebbian leraning rule during the whole life of the individual. 3. Then the fitness of the individual is calculated by the NIOS processor and sent back to the PC. --- Quote End --- Whether or not this maps nicely onto an FPGA depends on what the algorithm looks like. Would this algorithm be classified as requiring lots of parallel logic, for example, have people implemented the algorithm on GPUs? Do you have the algorithm working on desktop machines? What is the limitation there? Does slicing the processing up into lots of small parallel pieces help? Is there a lot of decision logic required? Cheers, Dave