PRINCIPAL COMPONENT ANALYSIS FOR MACHINE LEARNING
Polina Shulpina, V. A. Dokuchaev
Abstract
Training a Supervised Machine Learning model involves several stages. In the first stage, the data is passed via model, creating predictions (forecasts). The next stage is to compare these forecasts with factual values (ground truth). The final stage is optimizing the model by minimizing a certain cost function. The model improves that way. Occasionally, an input sample contains many columns. Using each column in a model leads to problems, the curse of dimensionality. At that rate, it is necessary to be selective about functions. We will embrace Principal Component Analysis (PCA), that is one of the main ways to reduce the dimensionality of data, losing the least amount of information.
Topics & Concepts
Curse of dimensionalityPrincipal component analysisComputer scienceArtificial intelligenceMachine learningColumn (typography)Component (thermodynamics)Function (biology)Sample (material)Principal (computer security)Stage (stratigraphy)Dimensionality reductionData miningBiologyFrame (networking)PaleontologyPhysicsEvolutionary biologyTelecommunicationsChemistryThermodynamicsOperating systemChromatographyAdvanced Data Processing TechniquesNeural Networks and Applications