How it Works
The Kohonen Self-Organizing Map simulates how the motor cortex develops its topographic map. Each neuron in the grid has a 2D weight vector representing a position in body space. Training inputs sample body positions, with the hand region sampled more frequently (simulating higher dexterity demand).
At each step, the best-matching unit (BMU) is found, then its weights and neighbors' weights are nudged toward the input. Over thousands of iterations, the map self-organizes so that nearby neurons respond to nearby body positions — forming a topographic map. The hand region expands to occupy more cortical area.
Δw_i = η · h(||r_i − r_BMU||, σ) · (x − w_i)
h(d, σ) = exp(−d²/(2σ²)) [Gaussian neighborhood]
η(t) = η₀ · exp(−t/T) [decaying learning rate]
Frequently Asked Questions
What is the motor cortex?
The motor cortex (primary motor cortex, M1) is the region of the cerebral cortex in the frontal lobe responsible for voluntary movement. It contains giant Betz cells whose axons form the corticospinal tract, directly controlling spinal motor neurons that activate muscles.
What is the motor homunculus?
The motor homunculus is a distorted representation of the human body mapped onto the motor cortex. Body parts with greater dexterity and finer motor control (hands, face, lips) have disproportionately large cortical representations, while less dexterous areas (trunk, legs) have smaller ones.
What is a Kohonen Self-Organizing Map (SOM)?
A Kohonen SOM is an unsupervised neural network that learns a low-dimensional topographic representation of high-dimensional input data. Neurons are arranged in a grid; during training, the best-matching unit (BMU) and its neighbors are updated toward each input, creating a continuous map.
How does the SOM algorithm work?
1) Present a random input vector. 2) Find the BMU (neuron with weights closest to input). 3) Update BMU and neighbors: Δw = η · h(d) · (x − w), where η is learning rate and h(d) is a Gaussian neighborhood function. 4) Decrease η and neighborhood radius over time. Repeat until convergence.
Why is cortical representation proportional to dexterity?
Brain areas controlling fine motor movements (hands, mouth) require more precise neural control, encoded in more cortical neurons. During development and with practice, active body parts expand their cortical territory through use-dependent plasticity — the more you use a body part, the more cortex it claims.
Can the motor cortex map change in adults?
Yes, cortical maps are plastic throughout life. In musicians, the representation of the playing hand enlarges with practice. After limb amputation, cortical areas from adjacent body parts expand into the deafferented zone. This cortical reorganization can also produce phantom limb sensations.
What is somatotopy?
Somatotopy is the organized spatial representation of the body surface in the brain. In the motor cortex, adjacent body parts are represented in adjacent cortical areas. This topographic organization is a fundamental principle of cortical mapping, also found in the somatosensory and visual cortex.
What is the difference between M1 and the premotor cortex?
M1 (primary motor cortex, Brodmann area 4) executes movements via direct corticospinal connections. The premotor cortex (area 6) plans and prepares movements based on sensory input and instructions. The supplementary motor area (SMA) is involved in internally-generated sequences and bimanual coordination.
What is cortical magnification factor?
The cortical magnification factor (CMF) describes how much cortical surface area is devoted per unit of body surface. High CMF (fingertips, fovea) means more cortical neurons encode a small area, enabling fine discrimination. The SOM simulation shows how differential input frequency produces differential map area.
How is the SOM related to cortical development?
The SOM is a computational model of how activity-dependent competition during development shapes cortical maps. Correlated activity from nearby body parts drives nearby neurons to respond similarly. This explains how the ordered somatotopic map emerges from initially disordered connections.