Machine Learning for Developers

Pros and cons of the analytical approach

The approach of using linear algebra techniques to calculate the minimum error solution is an easier one, given the fact that we can give a really simple representation that is deterministic, so there is no additional guessing involved after applying the operations.

But there are some possible problems with this approach:

• First, matrix inversion and multiplication are very computationally intensive operations. They typically have a lower bound of approximately O(n²) to O(n³), so when the number of samples increases, the problem can become intractable.
• Additionally, and depending on the implementation, this direct approach could also have limited accuracy, because we can normally use the limits of ...