Alternative way to use fitcsvm for code generation

how to use alternate syntax if fitcsvm is not supported for code generation

[Manikanta Aditya]

Hey,

The 'fitcsvm' function in MATLAB is used for training a support vector machine (SVM) for binary classification. However, if 'fitcsvm' is not supported for code generation, you can use the lower-level functions that fitcsvm uses internally.

Here’s an example of how you can do this:

% Assume X is your N-by-P matrix of predictors, and Y is your N-by-1 vector of responses
X = 10; % your predictor matrix
Y = 20; % your response vector

% Convert the responses to +1 and -1 for the SVM
Y(Y == 0) = -1;

% Set up the SVM problem
H = 0.5 * (X' * X);
f = -Y;
A = -diag(Y) * X;
b = -ones(size(Y));
lb = zeros(size(X, 2), 1);

% Solve the SVM problem using quadprog
alpha = quadprog(H, f, A, b, [], [], lb);

Minimum found that satisfies the constraints.

Optimization completed because the objective function is non-decreasing in 
feasible directions, to within the value of the optimality tolerance,
and constraints are satisfied to within the value of the constraint tolerance.

% Compute the bias term
bias = mean(Y - X * (X' * alpha));

% Now you can classify new data using sign(Xnew * (X' * alpha) + bias)

This code does essentially the same thing as fitcsvm, but uses 'quadprog' to solve the SVM problem directly.

Thanks!

[Keshav]

Thanks, Manikanta!

Answer 1

[Suman]

Hi Keshav,

You can try to use the "coder.extrinsic" function to use functions that are not supported for code generation. Please refer to this documentation to learn how to use it and the limitations associated with it: https://www.mathworks.com/help/coder/ref/coder.extrinsic.html