Quote:
Originally Posted by samihaq
1. How to use 1step learning for linear regression uses the term Ax + b = y,(b as a constant) as i think in the original form (w = pseudo_inv(X) * y ) it is used for Ax = y. We can take b on other side and subtract from y but then what values of b constant to use.

The constant
can be treated as a coefficient of
(which is the constant 1).
Quote:
2. Secondly when we have h(x) = b as hypothesis set then as it is just a constant and there are no parameters, so we cant use any learning algorithm then does it mean that we just to try a range of values for b ?

It is a constant in the sense that the value of the hypothesis does not change for different values of
, but the value of the constant itself can change from one hypothesis to another, so
should be treated as a parameter.