How did a Finnish professor teach neural networks to organize data into maps - without any labels at all?

The answer

In 1982, Finnish academician Teuvo Kohonen at the Helsinki University of Technology introduced the Self-Organizing Map (SOM), a type of artificial neural network trained using unsupervised learning to produce a low-dimensional (typically 2D) representation of high-dimensional input data. The key innovation is topology preservation: data points that are similar in the input space end up mapped to nearby neurons on the grid, creating an intuitive visual 'landscape' of the data. SOMs achieve this through competitive learning - for each input, the closest neuron (the 'best matching unit') and its neighbors adjust their weights toward the input vector. Kohonen SOMs have been applied to problems ranging from speech recognition and financial analysis to genomics, and the original 1982 paper has garnered over 20,000 citations, making it one of the most influential works in unsupervised learning.