News for hackers. Stuff that matters.
Re:How UMAP Works by optimalsolver on Saturday July 03, 2021 @11:31AM (#27722262)
So the reduced features can be used for inference/prediction?

Also, is there some rule for choosing a) the number of dimensions the unreduced dataset should be mapped on to, b) the number of neighbors?

I assume the default parameters would work for most tasks.

Re:How UMAP Works by paulgb on Saturday July 03, 2021 @12:34PM (#27722817)
UMAP is typically used for embeddings that get displayed to a human, so 2 or 3 dimensions is typical.

It doesn’t really suit inference/prediction because you can’t really add new data without influencing the embedding values of the other data. It’s not like PCA where you can learn a projection once and then map new data points to the same embedding space.

Re:How UMAP Works by nope_42 on Saturday July 03, 2021 @12:40PM (#27722871)
Re:How UMAP Works by nestorD on Saturday July 03, 2021 @01:13PM (#27723121)
You can but it is both costly and a hack (the resulting embedding will not be as good as the one you would have gotten restarting from scratch). So I would not recommend using it in an inference pipeline.
Re:How UMAP Works by lmcinnes on Saturday July 03, 2021 @06:17PM (#27725396)
If this is a thing you want to be able to do efficiently then ParametricUMAP (see [docs](https://umap-learn.readthedocs.io/en/latest/parametric_umap....) and [the paper](https://arxiv.org/abs/2009.12981)) will be very effective. It uses a neural network to learn a mapping directly from data to embedding space using a UMAP loss. Pushing new data through is only slightly more expensive than PCA, so being part of an inference pipeline is fine.
Re:How UMAP Works by nestorD on Saturday July 03, 2021 @11:28PM (#27727011)
But then you have to train a neural network and lose on the speed advantage of UMAP (offline yes but still much slower and finicky).
Re:How UMAP Works by lmcinnes on Sunday July 04, 2021 @10:37AM (#27729927)
It is really not that much slower for training (see the paper), and if you are interested in pipelines the difference is not so great considering you are looking at a one off training time vs. lots of inference.