Question regarding Neighborhood identification

[Shawnmhy]

Hi team, great work on this amazing project. I was trying to translate the idea of identifying neighborhoods using MiniBatchKMeans clustering to my own dataset. I understand that this function can cluster cell frequency vector into designated number of clusters, say N = 10. However, I am having trouble understanding how the final clustering heatmap can be drawn. Since I would imagine the number of rows of the the heatmap should be corresponding to the number of cells being clustered. One possible solution I can think of is calculating the mean for each cluster, however, the legend of the heatmap ranges from -3 to 3, and I don't understand why a negative value can be derived. Please help me through, thank you!