I think Nu should be written instead of Mu according to the formula what gives us Mu prime.

Original article can be found here (source): Deep Learning on Medium

2. Then, the new mean, Mu prime, is the weighted sum of the old means. The Mu is weighted by r-square, Nu (instead of Mu in the article) is weighted by Sigma square, normalized by the sum of the weighting factors. The new variance term would be Sigma square prime.