different results for same data in R 3.4.2 vs. R 3.5.3

[pmainker]

Hi I executed tsne using same parameters in RStudio [3.4.2] and Alteryx R module [3.5.3] and got very different results. Any thoughts? Both have same tsne package version. no other packages are involved in data manipulation, it is a straight read in, apply tsNE,write out.

[wleoncio]

If you're leaving initial_config = NULL (the default), its values are randomly drawn from a normal distribution. I'm not familiar with Alteryx R modules, but R 3.5.0 introduced some changes to its random number generator which could explain the differring results (more info here). In any case, the results should be stochastically equivalent.