The human brain consists of a highly complex network of approximately 85 billion connected neurons, which continually exchange information with each other. In order for this complex network to function efficiently, it is important that the distances between neurons encoding similar properties remain relatively short.
“We realized that all of the various conformations of the neuron and its
various components are simply morphological adaptations governed by
laws of conservation for time, space, and material”
(RAMÓN Y CAJAL)
Distinct regions of our brain are responsible for different tasks, such as vision, language, and memory. Within these regions, neurons that respond to similar features are also located near each other, forming so-called neural maps. If similar cells are in close proximity to each other, the paths connecting them are shorter leading to a more efficient layout that saves wiring cost and speeds up signal times.
“If cortical maps are selected in the course of evolution to improve the
fitness of the organism, they can only be chosen on the basis of the length
of connections”
(DMITRI ‘MITYA’ CHKLOVSKII)
If the relative position of a neuron depends on its connection to other neurons, models that predict layouts which follow the general design principles of neuronal networks can be created. The MAPS toolbox exploits common dimension reduction methods like multidimensional scaling (MDS) and t-distributed stochastic neighbour embedding (t-SNE) to predict neuronal layouts by the connection dissimilarity of neurons.
Using the connection dissimilarity of neurons with these methods ensures that neurons which share a similar connectivity are placed more closely to each other than neurons with a less similar connectivity. Finally saving wiring length compared to layouts where the most interconnected neurons would be placed more far apart.
In the human visual system and in that of many mammals, the neurons that respond to objects with similar orientation are indeed located near each other.
