[osoba]

For Calculus check Coursera - the Ohio State course if you really don't know absolutelly anything, then do the Penn State course(s) Calculus in a Single Variable

For Linear Algebra check Gilbert Strang's course on MIT OCW

For Probability/Statistics either John Tsitsiklis' course on MIT OCW (or even better on edX when it re-runs) or Harvard's Statistics 110: Probability course on youtube

The level of multivar calculus you'll need is really minimal, just google these terms "partial derivative", "total derivative", "differentials and integrals commute" ("Symmetry of second derivatives", "Fubini's theorem") or take the vector calculus course from MIT OCW

[jules]

Indeed you don't need to know much vector calculus for machine learning. Stokes theorem and the like are mainly relevant for physics, but the Jacobian and Hessian are good to know.