As IoT progresses from simple devices feeding data to the cloud for analysis to smart devices performing sophisticated inferencing and pattern-matching, the importance of heterogeneous computing becomes even more paramount. Bringing AI to the edge provides a plethora of benefits, including quicker response, enhanced security, improved mobility, and lower communications cost. Since particular combinations of neural networks and frameworks running on specialized computing cores are ideal for specific tasks, heterogeneous computing is the best strategy for deploying AI. In addition, AI design challenges (accuracy, responsiveness, SWaP, and cost) can be met by adopting a heterogeneous computing approach.
This brief discusses the tradeoffs for these core types when implementing AI “at the edge.” In addition, it covers the techniques ADLINK uses to help its customers optimize their AI solutions.