I'm using Support Vector Machine(SVM) with image classification. Each SVM model results a linear model
$$y = wx + b$$
Where $w$ and $b$ is the SVM parameters.
If I have multiple SVM models, I will get a vector $Y$ and matrix $W$ and vector $B$
$$Y = Wx + B$$
In my current case, I have a activation function called sign
$$Y = sign(Wx + B)$$
Now, $Y$ can holds values such as 1 or -1 only. In my case, $Y$ has $n$ dimensions (rows) and therefore $W$ and $B$ are a $n$-class SVM model. The index of $Y$ that holds the number 1, is the class ID of vector $x$.
The problem is that what if I get two number 1 inside vector $Y$ like $Y = [-1, -1, 1, -1, -1, -1, 1, -1, -1, -1, -1]$ ?
I can solve this by using propability. But my data is small, about 150-200 $Y$ vectors.
Assume that we are removing the activation function sign and using the pure data of $Y$ instead. Which method sould I use here to determine which $Y_i$ for a given $x_i$, is giving the user the class ID of vector $x_i$?
- Decision trees (multiple if-statements)
- Naive Bayes Classifier
- Markov chains
- Linear regression
The goal here is to find an integer number as class ID for vector $x$.