I will survey here the recent rich line of works on the statistical problem of Tensor PCA. How hard is it to denoise a (small-rank) tensor in high dimension? I will discuss the natural thresholds obtained from the point of view of Information Theory, Statistics and Optimization.
62F10