Observation weights

[ilyakorsunsky]

trafficstars

Would it be possible to assign weights to observations, so that the user can define how much a single observation contributes to the overall objective function? For my particular use case, I have largely imbalanced datasets, with 10X more samples in than the other. With regular tSNE, most of the real estate is taken up by the larger dataset. I would like a representation that gives equal space to both datasets.

Thanks for your feedback and for a wonderful tool!

[dkobak]

This is not currently possible in FIt-SNE and I have never seen this done in the literature. Sounds more like a research project.