Week 2: Question regarding Bias Correction

[Marcia_Ma]

In video 5, Professor Andrew Ng wrote the new V2 with bias correction as (0.0196 * theta1 + 0.02 * theta2) / 0.0396. I was wondering why the numerator contains the term 0.0196 * theta1 rather than 0.98 * theta1? If we also implement bias correction for V1, then we should get V1/0.02 = 0.02 * theta1/0.02 = theta1.

Thanks in advance!

[paulinpaloalto]

It’s 0.0396 and that’s because it’s (1 - \beta^t) where \beta = 0.98 and t = 2.

0.98^2 = 0.9604
1 - 0.9604 = 0.0396