Sum-of-Squares Lower Bounds for Sparse PCA
This provides a hardness result for a powerful family of convex relaxation algorithms in high-dimensional machine learning, but it is incremental as it extends known lower bounds from degree-2 to degree-4.
The paper tackles the Sparse PCA problem by proving that degree-4 Sum-of-Squares convex relaxations cannot improve the quadratic sample complexity gap, showing they require n ≈ k² samples like basic methods, whereas information-theoretic limits allow n ≈ k log p.
This paper establishes a statistical versus computational trade-off for solving a basic high-dimensional machine learning problem via a basic convex relaxation method. Specifically, we consider the {\em Sparse Principal Component Analysis} (Sparse PCA) problem, and the family of {\em Sum-of-Squares} (SoS, aka Lasserre/Parillo) convex relaxations. It was well known that in large dimension $p$, a planted $k$-sparse unit vector can be {\em in principle} detected using only $n \approx k\log p$ (Gaussian or Bernoulli) samples, but all {\em efficient} (polynomial time) algorithms known require $n \approx k^2$ samples. It was also known that this quadratic gap cannot be improved by the the most basic {\em semi-definite} (SDP, aka spectral) relaxation, equivalent to a degree-2 SoS algorithms. Here we prove that also degree-4 SoS algorithms cannot improve this quadratic gap. This average-case lower bound adds to the small collection of hardness results in machine learning for this powerful family of convex relaxation algorithms. Moreover, our design of moments (or "pseudo-expectations") for this lower bound is quite different than previous lower bounds. Establishing lower bounds for higher degree SoS algorithms for remains a challenging problem.