2019年4月12日 · With UMAP, you should be able to interpret both the distances between / positions of points and clusters. Both algorithms are highly stochastic and very much dependent on choice of hyperparameters (t-SNE even more than UMAP) and can yield very different results in different runs, so your plot might obfuscate an information in the data that a ...

2019年3月21日 · I think UMAP is very promising and is a great contribution but to be honest I am getting a little bit annoyed with all the marketing and the hype that surrounds it. People think that t-SNE cannot embed new points but UMAP miraculously can. In reality, t-SNE can do it just as well as UMAP can; it is just a matter of convenient implementation.

$\begingroup$ @amoeba there is no doubt that tSNE produces nice visualizations. But it can also produce "false" clouds. In the Shekar example, they did not cluster on the tSNE projection as far as I can tell; they only used tSNE to visualize. That is okay, and you can spot a few objects clustered differently than a post-tSNE clustering would do.

2020年12月26日 · On the other hand, t-SNE and UMAP are non-linear visualization tools. They allow to visualize data in a lower dimensional space withouth losing so much information and without assuming any linearity. Thus, this basically means that your data is non-linear.

A recent preprint provides some intuition on this topic, comparing t-SNE, UMAP, Force-directed graphs (FDGs; the ForceAtlas2 FA2 implementation), and diffusion maps (Laplacian eigenmaps here). In particular they show that under certain parameterisations that change "the balance between the attractive and the repulsive forces" t-SNE embeddings ...

2020年12月26日 · Hi @dantferno, from my point of view it's usual that UMAP works better than t-SNE just because of their mathematical roots. If you're interested in that you can go to the papers and check it, but to sum up: UMAP is a enhanced version of t-SNE. I always use UMAP over t-SNE, I didn't find yet a single case in which t-SNE works better than UMAP.

2022年4月26日 · According to the UMAP paper: Our algorithm is competitive with t-SNE for visualization quality and arguably preserves more of the global structure with superior run time performance. paper; It seems to me that UMAP is better when it comes to visualizing. However is there a reason for choosing t-SNE over UMAP?

2016年10月5日 · In some fields of biology we are dealing with highly dimensional data where t-SNE simply does not scale. Therefore, we use PCA first to reduce the dimensionality of the data and then, taking the top principle components, we apply t-SNE (or a similar non-linear dimensionality reduction approach like UMAP) for visualisation.

2021年6月19日 · PCA and UMAP give functions that can be applied to new data. In that sense, you can learn a PCA or UMAP transformation for several folds and then apply that transformation to new data. Unfortunately, t-SNE does not work this way; t-SNE works on all data at once and does not produce a function to apply to new data.

2024年11月30日 · The question is the same as posed in the title; is there an alternative to PCA that doesn't rely on the linear assumption but maintains distances (i.e. the main issue with UMAP/tSNE)? Thanks! pca