#classifier-free-guidance

Classifier-free guidance combines the conditional and unconditional noise predictions of one network as $\tilde\epsilon = (1+w)\,\epsilon_\theta(z, c) - w\,\epsilon_\theta(z)$. At one pixel, $\epsilon_\theta(z, c) = 0.6$ and $\epsilon_\theta(z) = 0.2$. What is $\tilde\epsilon$ for the guidance strength $w = 3$?