Dimensionality Reduction and Clustering

Supervised learningSupervised learning approaches discussed thus far, classification and regression, rely on learning a mapping between the input features and the output labels based on a ground truth data. This approach inherently assumes a label associated with each datapoint, which needs to be learnt. However, in many situations there might not exist a label associated with the data, while there might be several features. The goal in this case would be to group “similar” datapoints together or to identify those minimal features that represent the data in a meaningful fashion. Such problems can be handled by unsupervised ML algorithms such as clustering. In this chapter, we will discuss various clustering algorithms. We will discuss how dimensionality reduction can be achieved by unsupervised approaches such as principal component analysis. We will also discuss algorithms such as k-means, Gaussian mixture modelModels, and t-SNE.