Look at this J funciton: \(J(\vec{w},b)=\frac{1}{2m}\sum^m_{i=1}(f_{\vec{m},b}\vec({x}^{(i)})-y^{(i)})^2+\frac{\lambda}{2m}\sum^n_{j=1}w^2_j\)

it makes unacceptably large errors in predictions.

What do you try next?

Get more training examples

Fixes high variance (or we could say overfitting)

Try smaller sets of features

Fixes high variance

Try getting additional features

Fixes high bias(or we could say our modal is not so accurate)

Try adding polynomial features($x_1^2,x_2^2,x_1x_2 etc$)

Fixes high bias

Try decreasing $\lambda$

Fixes high bias

Try increasing $\lambda$

Fix high variance