Impact of metrics on the effectiveness of Kohonen network clustering

The research paper focuses on investigating the impact of different metrics on the clustering process in Kohonen networks, also known as self-organized maps (SOM). The theoretical foundations of Kohonen networks, including their structure and algorithm of operation, are presented. Various metrics, such as Euclidean distance, Manhattan and cosine distance and their potential impact on clustering process are then discussed. Experiments were conducted using the Iris dataset constrained to two dimensions using PCA.