:hammer: DEVoTED

- 1 min

Project on a clinical decision support system for an early estimate of treatment resistance in depression.
Please find all the relevant information and our cooperation partner here!

Methods: Self-Organizing Maps (SOMs)

In machine learning terminology, SOMs belong to the categories of artificial neural networks and unsupervised learning procedures. In contrast to other neural networks, SOMs are based on competitive learning mechanisms instead of error reduction (such as backpropagation in many supervised learning tasks). They are also known as Kohonen maps as they were first introduced by Teuvo Kohonen (1982).

SOMs are usually 2-dimensional representations of data which use neighborhood functions to incorporate topological properties of the input data. This way, they are well suited for visualizing high dimensional data in 2-D space. A SOM consists of a 2-D grid of nodes, i.e. weight vectors, arranged in a rectangular or hexagonal shape. Each node represents a position in the data space and thus has the same dimension as each input vector (the data samples).

How the algorithm works - the SOM pipeline:

  1. The weights of each node are initialized (with neighboring nodes having similar weights).
  2. A random vector (sample) is selected from the input data (e.g. the training data set).
  3. The vector is compared to every node to evaluate which node’s weights are most similar to the input vector. The winning node is called the Best Matching Unit (BMU).
  4. The winning node’s weights are changed towards the values of the input vector. The neighbors also become more similar to the input vector. The closer on the map a node is to the BMU, the more its weights get altered. Nodes that are further away, change their values less.
  5. Repeat steps 2-4 for N iterations.

Thus, by shifting the nodes’ weights towards the input sample values, the SOM becomes a more and more precise representation of the potentially high dimensional input data. Data samples with similar values will have the same BMU and will be placed in the same node or in a neighboring node on the SOM.

References:

rss facebook twitter github gitlab youtube mail spotify lastfm instagram linkedin google google-plus pinterest medium vimeo stackoverflow reddit quora quora gscholar