Performance comparison LASSO vs. Tikhonov

128 Views Asked by At

Tikhonov Regularization can be described as

$$ \min_g \| Yg - x\|_2^2 + \alpha \|g\|_2^2 $$

and LASSO can be described as:

$$ \min_g \| Yg - x\|_2^2 + \beta \|g\|_1 $$

Is there any way to relate the resulting error (e.g. mean squared error) between these two?

The answer depends on $Y$, $x$, $\alpha$ and $\beta$ but is there anything that can be said if either of these have certain forms?