Clustering in R | Hacker News Reader

Clustering in R

blog.dominodatalab.com

57 pointsgk16 comments

Glad to see the article mention dropping in the fastcluster package to replace the default R hclust. I would suggest using the parallelDist library as well instead of the standard dist.

Clustering in general and hierarchical clustering have been something I have spent some recent time trying to come up to speed on. Current state of the art seems to be using graph-based approaches of community detection (Louvain method) where graphs are built for a set of samples with N features by starting with a graph of K-nearest-neighbors and assigning some weight to the edges.

> Current state of the art seems to be using graph-based

Do you know how this compares to UMAP?

I strongly suspect that it's worse than UMAP or Ivis. Graph based methods are great for some things but not clustering

We've found that graph-based community approaches have some really nice benefits in our bioinformatics data.

In particular, we have found that these approaches seem to preserve very small cluster structure "better" than traditional approaches. Meaning, we have a small group of cells that we know belong to their own cluster group and the graph-based community approaches preserve these "small" groups outside of other clusters nicely.

But we have also noticed (and had some feedback) that we windup with final modularity scores that are very high - greater than or equal to 0.90 (on a scale of -1 to 1). Applied math folks in the graph algorithms world kind of seem to look at that and go "eh, that is so high you should probably just do PCA and move on . . . "

Especially given that you could (and people seem to) use UMAP as a precursor to louvain methods, I'll probably be looking into UMAP to see how it goes. But our current performance bottleneck is that the clustering (or community approach of graphs with the louvain method) is our computational bottleneck, so we'd like to whittle that runtime down as much as possible.

Nope. But I think UMAP would be a dimensionality reduction step that you fed into a graph-based community extraction algorithm? In our current code experiments, we have been starting with a k-nearest-neighbor approximation algorithm to build the graph. UMAP would, at first glance, replace that.

I think there is even a louvain clustering method (really a graph-based community extraction method) built into several UMAP libraries floating around . . .

What a great article. Great visualizations, and I particularly like the focus on post clustering analysis. This doesn't get addressed enough in undergraduate/ graduate level training on clustering.

I will raise one 'issue', that I expected to see in the first section regarding k-means, and clustering in general, which is the issue of variable importance/ variable selection.

I didn't see anything addressing normalization of values, how to deal with factor variables, or variable weighting. Clustering for data science is something that I engage in daily and while its a bit of a detail, how you normalize (if you do so), and how you weight variables is the key to getting a useful set of clusters or not. Maybe its an aside, but maybe step through the k-means algorithm early on and show why normalization and weights matter. Just a suggestion.

Edit: I suppose this is addressed through PCA. There are ways of addressing these issues outside of PCA.

Reply on news.ycombinator.com