The basic MatrixMath library is okay. If you want the high-level simplicity of Matlab/Octave for intense linear algebra operations, the Eigen3 C++ library is almost just a simple yet much faster.
Several projects use it including Google, the European Space Agency, some mobile apps, etc.
I have actually tested Eigne3 on the Arduino Due, and it works. Here's an excerpt from an Arduino sketch demonstrating Matlab/Octave-like programming, on the Due:
// Kalman Gain Example
// Matlab form: K = Pp * H' * inv(H * Pp * H' + R)
// On Arduino:
X = H * Pp * H.transpose() + R;
K = Pp * H.transpose() * X.inverse();
I detailed out instructions for obtaining and setting up Eigen3 for the Arduino Due IDE here: