Some neuroscientists have suggested that neurons actively seek to have variability in their responses over time. This may make for more robust neural circuits that can tolerate variation in their inputs, such as from noisy neural sensors. Neural noise analysis is a growing area in systems neuroscience.
Complete genetic coding outlining the entire structure of the nervous system is possible for an animal, like an invertebrate, that has a few hundreds or thousands of neurons. In fact, in many invertebrates, every neuron and the majority of synapses are genetically specified and identical from animal to animal. But in vertebrates, which have millions or billions of neurons, such a development scheme isn’t feasible via coding by about 20,000 genes.
Therefore, in vertebrates, the nervous system is developed by a process that specifies general rules for connectivity but enables stimuli from the environment and random events within the developing nervous system compete for synapses. In other words, important aspects of synaptic and neural fitness are locally determined within the framework of the genetic master design. This is certainly a valuable construction procedure for a system that has to build itself.
One of the principles that allows this system to work is homeostasis, in which each neuron maintains a certain balance of excitatory and inhibitory inputs, and a minimum number of active output synapses in order to survive; otherwise, it dies. Competition for a limited number of input synapses is a mechanism that makes neurons in a given area respond to different input combinations, including new ones that are acquired during learning.