4: SVM Flashcards
Is SVM SL or UL?
SL
In what kind of domain is it used?
In various domains that require high discrimination power & accurate classification and prediction algorithms. Moreover, this algorithm can operate on both continuous and categorical data and can solve both linear and nonlinear problems. (imagine classification, text categorization, face detection, sentiment analysis, hand writing recognition)
What is SVM about?
A mathematical model within the class of supervised learning for:
accurate classification & prediction
both continuous & categorical data
both linear & non-linear problems
high efficiency & effectivity
When is the hyperplane learned?
It is detected during the training phase. The SVM algorithm uses the training instances to construct the linear function of the hyperplane. The goal is to use that description of the hyperplane to construct a function that maps each training instance x to a function value f(x) for which holds f(x) > 0 for instances of one class, and f(x) < 0 for members of the other class.
How does SVM work?
- The position of the hyperplane is derived in the training phase of the SVM algorithm.
- In the testing phase, the new non-labeled instance is classified as class according to its location relative to the hyperplane.
The hyperplane above is represented as a straight line that passes through the middle of a gap. Where does the gap lie?
This gap lies between the closest instances of the different classes.
What is the goal of SVM?
The goal of the SVM algorithm is to detect (or learn) the hyperplane that maximizes the width of the margin. The width of this margin is measured based on the perpendicular distance from the separating hyperplane to the support vectors.
What are the support vectors?
The data points that determine the margin.
What is the margin?
The boundary gap between two point-sets.
What is the hyperplane?
A separating boundary that helps to classify the points. The hyperplane in SVM forms a space that separates the instances of different classes. This hyperplane is a straight line if the space is two-dimensional, and is a plane if the space is three-dimensional, and so on. In other words, in n-dimensional space, the hyperplane constitutes an -dimensional subspace.
What is a hyperplane?
The SVM algorithm is a supervised classifier as it uses labeled instances in training to learn how to classify new non-labeled instances.The SVM classifier finds the maximum margin hyperplane in the training data to classify non-labeled instances in the testing phase.The hyperplane maximizes the margin between data points of differing non-classes.
What are the different methonds in SVM?
- classification hard margin - classification soft margin primal form - classification soft margin dual form - regression (SVR)
SVM Algorithm:
- The separating hyperplane is detected (learned) by maximizing the margin width 2. The margin width is measured based on the perpendicular distance from the separating hyperplane to the support vectors
- Large margin ensures good generalization
- This method is memory efficient, as it uses only a subset of training points (support vectors)
How is the HARD margin found?
Training the model without training errors.
What is required to determine the linear equation of a hyperplane in machine learning?
The determination of the hyperplane linear equation requires the calculation of the weight vector w and the displacement value b, such that the margin between the support vectors is maximized by this specific hyperplane. However, it is not always possible to find a hyperplane {represented by w and b }, where all instances are satisfying this optimization formula (hard margin). It pre-supposes that all training instances are on the correct side of the hyperplane, and this case is often not feasible in real-life data.
