Skip to main content

Advertisement

Fig. 1 | BMC Bioinformatics

Fig. 1

From: GARS: Genetic Algorithm for the identification of a Robust Subset of features in high-dimensional datasets

Fig. 1

Block diagram of the GARS workflow. The first population of chromosomes (red block) is created by randomly selecting sets of variables (see the red box on the left). Then, each chromosome is assessed (green block). To do this (see green box on the left), we designed a fitness function that (A) extracts for each sample the values of the variables corresponding to the chromosome features, (B) uses them to perform a Multi-Dimensional Scaling (MDS) of the samples, and (C) evaluates the resulting clustering by the average Silhouette Index (aSI). Finally, to obtain a new evolved population, the Selection (light blue block), Reproduction (blue) and Mutation (purple) steps are implemented. This process, iteratively repeated several time, allows to reach the optimal solution. f = feature, s = sample, v = value of a features in a sample, n = total number of samples, m = total number of features, rnd (1,m) = random integer between 1 and m, i = specific sample, a(i) = average dissimilarity of i with respect to all other samples within the same class, b(i) = the lowest averaged distance of i to all samples belonging to any other class, aSI = average Silhouette Index, and MDS = Multi-Dimensional Scaling

Back to article page