t-SNE summarization

Part 1 (2017)
1

T-SNE is generally used to see if there are any patterns in the dataset. Or even as the dimentionality reduction to see how the data is like.
Reading this blog, http://distill.pub/2016/misread-tsne/
How can I be sure that which of the hyperparameters is giving the correct result. Since I have no way to know what the pattern of the complete dataset looks like, I have no way of distinguishing between correct and wrong.

Kindly advice.

2

There isn’t any notion of ‘correct’ or ‘wrong’ for t-SNE, only a subjective ‘better’ or ‘worse’. Choosing parameters for t-SNE requires you to think carefully about your problem and then search across different parameter settings.

Remember that it is only a technique for visualisation, so it is largely up to you to decide whether the output looks good or not.