Should not the theta be in the range of function?

#1

Sir, when we talk about taking any random theta shouldnt the theta be satisfying the curves equation?
If yes then how would we know what theta to take if we do not know the curve and training data?
And if not then how will we calculate the gradient of curve at theta?

#2

Hey @Kunnaal, first of all taking when there is only one theta, on y axis we have loss, and there is a value of loss corresponding to all theta values so no matter which value of theta you imagine, there is a corresponding loss value and ultimately there will be a point on the curve.

I think you misunderstood that we have one theta on x axis and other on y axis, this is not the case. On y axis we have loss. And for any values of theta there exist a loss value which is on the curve itself.

Hope this resolved your doubt,
Plz mark the doubt as resolved in my doubts section. :blush:

#3

Okay thanks
and what about the x that we consider which afterwards decreases to the local minima?
It has any restrictions?

#4

Hey @Kunnaal, no there are no restrictions on theta ( which you referred as x).

Hope this resolved your doubt,
Plz mark the doubt as resolved in my doubts section. :blush:

1 Like
#5

ok thanks
:smile:
i got it