Branch Specialization

While the parallels between branch specialization in artificial neural networks and neural circuits in the brain are striking, there are clearly major differences and many outstanding questions. From the perspective of building artificial neural networks, we wonder if branch specific tuning of individual units and their connectivity rules would enhance performance? In the brain, there is good evidence that the activation functions of individual neurons are fine-tuned between and even within distinct neural circuits. If this fine tuning confers benefits to the brain then we might expect similar benefits in artificial networks. From the perspective of understanding the brain, we wonder whether branch specialisation could help make experimentally testable predictions?