There are lots of discussions around it that you can google (e.g., https://www.kaggle.com/code/agsam23/pca-vs-t-sne), but it boils down to the fact that T-sne is a non-linear transformation of the data and ultimately not useful or appropriate to construct morphospaces (to be a morphospace as need to be able to see same amount of change per unit component change)
Otherwise if your goal is simply cluster your data without the care for preserving the Euclidean distances in your projected space (such as they do in single-cell RNA or other methods), sure you can use T-SNE. On Monday, June 12, 2023 at 4:24:41 PM UTC-7 Juan Esteban Vrdoljak wrote: > Hi all! > > I'm currently exploring alternatives to PCA, CVA, and CMD for data > visualization. A colleague introduced me to t-SNE (t-Distributed Stochastic > Neighbor Embedding, Van der Maaten & Hinton, 2008), which is a method used > to reduce high-dimensional data. This is the first time I've come across > this method, and I haven't seen it mentioned in many "theoretical" articles > on geometric morphometrics (except for Courtenay, 2022). > > I would appreciate hearing your opinion on the application of t-SNE in > geometric morphometrics. What are its advantages, disadvantages, and > potential uses? > > Thank you so much for your input, and I apologize for asking such an > open-ended question. > > > > Van der Maaten, L., & Hinton, G. (2008). Visualizing data using t-SNE. > *Journal > of machine learning research*, *9*(11). > > Courtenay, L. A. (2023). Can we restore balance to geometric > morphometrics? A theoretical evaluation of how sample imbalance conditions > ordination and classification. *Evolutionary Biology*, *50*(1), 90-110. > -- You received this message because you are subscribed to the Google Groups "Morphmet" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/morphmet2/3b7ba739-2825-4ebc-ac54-b1f39bbd683en%40googlegroups.com.
