Question

Linear classifiers. Discuss the differences between LDA, Logistic Regression, separating hyperplanes, and the Optimal Soft-Margin Hyperplane. In what situations would you use each of the classifiers?

Answer 1

81 Logistic regression indeed does not assume any specific shapes of densities in the space of predictor Ivariables, but LDA does. | Discriminant analysis is really used only for a Categorization, logistic regression is often used when we aren't even interested in categorization but in getting the odds ratios for each variable.

PAGE! DATE: LDA: It assumes data is normally distributed, all groups are identically distributed in case the groups have different covariance matrices. LDA becomes aDA Isum: Support vector machine generalizes the optimally - seperating Hyperplane (OSH). OSH assumes that all groups are totally seperable. Svim makes use of la slack nerable that allows a certain amount of overlap between the groups. srm classification is an optimization problem DA has an analytical solution. soft - margin Hyperplanes: A perfectly seperating - hyperplane may not always exist in the input space for a desired margin width are could solve this by projecting the data into a higher. dimensional space using a polynomial or other non-linear function. To allow a few margin errors in the training set, we use slack variables. * situations: IDA LDA is dimension reduction technique which is - commonly used for the supervised classification problems. It is used for seperating two or more classes. It is used to project the features in higher dimension space into a lower dimension space. logistic regression is used to describe date and t explain the relationship between one dependent binar nariable and one or more nominal, ordinal interval or ratio dovel independent variables.

6 sums operate on principle of seperating hypesplanes for maximize margin and only daka points near margin are relevant (the support rectors) 9 Optional soft - margin hyperplane. It is really effective ono in higher dimensioni It is best algorithm if your data are seperable. only support rectors affect the optimally spaced hypeoplane, so it is less affected by outliers.