Computer Vision News - December 2020

Computer Vision Tool 12 The demo program is and after importing the needed packages, it defines four functions which are explained in the next paragraphs. The function closest_node() gives indices for the row and column for a SOM with size m_rows x m_cols . As explained earlier, those are the coordinates of the map cell, the vector of which is closest to the data item at data[t] . Best matching unit (BMU) is the nearest cell vector to a specified data item. The function euc_dist() is the Euclidean distance between two vectors, which is defined as the square root of their location difference, or by the Cartesian coordinates of the points using the Pythagorean theorem. The function most_common() provides the most common value, using as input a list of integer values. The variables in the section “0” hold the dimensionality of the dataset, and the number of rows and columns of the SOM. Range_max is themaximum Manhattan distance for any two cells in the SOM, lr_max is the initial learning rate used when constructing the SOM and steps_max specifies the number of training iterations to perform. isom.py Image 1: A 40x40 grid of neutrons (SOM) using Fishers Iris flower data set (250 iterations). A color image formed by first three dimensions of the four dimensional SOM weight vectors (top left), pseudo-color image of the magnitude of the SOM weight vectors (top right), U-Matrix (Euclidean distance between weight vectors of neighboring cells) of the SOM (bottom left) and overlay of data points (red: I. setosa, green: I. versicolor and blue: I. verginica) on the U-Matrix based on the minimum Euclidean distance between data vectors and SOM weight vectors (bottom right) Source: Wikipedia