Kernel Functions and their Applications
Kernel functions are the critical element that give SVMs their power, allowing them to handle non-linear data effectively. By mapping the original data into a higher-dimensional space, kernel functions enable SVMs to find linear decision boundaries in this new space, which correspond to non-linear boundaries in the original space – sounds confusing at first but think of it like an accordion: the accordion starts off compressed in a flat state analogous to our initial dataset’s feature space. When we expand the accordion, we go from a flat feature space to a three-dimensional space which spreads our data points out along this new dimension. We can use this new dimension to find our decision boundary and then squeeze our accordion back to its original flat state. The linear boundary we found when the accordion was stretched out will also collapse into a potentially non-linear (i.e., curved) state.
In this recipe, we will explore different...
