The goal is to reduce the cost. Not getting to 0 is fine.

If you run more steps then it will get closer to zero but in some cases it may end up increasing because of how gradient descent works.

    Soon Hin Khor, Ph.D.

    Written by

    Use tech to make the world more caring, and responsible. Nat. Univ. Singapore, Carnegie Mellon Univ, Univ. of Tokyo. IBM, 500Startups & Y-Combinator companies