How do Support Vector Machines (SVMs) differentiate between a glass and a bottle or between a malignant and a benign tumor when it dealing with it for the first time?

What will be the analysis mechanism involved in this?


I will try to give you a simplified explanation of how SVMs work.

The data one works with can be of two types. Either it is very easily separable and there is a clear straight line boundary between data points of different classes. We call such data linearly separable(image) or the data points of the classes are mixed in a way that there is no clear straight line boundary that separates them. In other words the data is not linearly separable.(image).

One way to make data points linearly separable is to map them into a higher dimension where they would become linearly. If we have a set of data points which are not linearly separable in 2 dimensions, we could map them into 3 dimensions where the data becomes linearly separable(like so).

SVMs work by converting data from lower to higher dimensions where it is linearly separable and then tries to find the boundary that separates the data.

The same concept could be applied to images, where the images are converted into a very high dimension in which images of different types can be separated linearly.

This could mean that the SVM model trained is able to easily distinguish between a bottle and a glass or different tumour types but that would depend entirely on the data used to train the model. The problem arises when the distribution of the training data and testing data is completely different. So for example, you trained the SVM to distinguish between glasses and cups of a certain shape and colour and while testing you show it glasses of completely different shapes and colours, the SVM will not be able to distinguish between them.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.