Support Vector Machines videos
Support vector machine finds the optimal hyperplane that separates data into different classes with the maximum margin for classification or regression tasks.
<aside> 💡 Useful for dealing with high-dimensional and nonlinear data.
</aside>
Consider a dataset that contains two classes: Orange & Green. A good "enough model" can easily classify them. We just need a single line between the classes. Easy. The real challenge in machine learning isn't just about classifying the current data. We need a model that works with unseen data as well.
But there is a twist:
If you look at the image. The 'good enough' model is heavily affected by an outlier. So what happens if new data comes in? We will have a lot of misclassification, thanks to the outlier.
Our "good enough" model will start to perform badly. The model may be unable to adapt and lack flexibility.
But we have SVM.
In SVM we still draw a line, but we also add a safe zone. This safe zone or margin, is the key to superior performance. In this safe zone, we allow some room for outliers.
With the safe zone, the model will handle new data more effectively.
With fancy words, SVM finds the perfect line with the maximum margin between classes with some room for error/misclassification.
The safe zone future-proofs the model and it remains more robust. When new data comes in an SVM is less likely to misclassify it.
Welche Ebene ist die Beste?