On the Connection Between Learning Two-Layers Neural Networks and Tensor Decomposition

We establish connections between the problem of learning a two-layer neural network and tensor decomposition. We consider a model with feature vectors boldsymbol x ∈ mathbb Rd, r hidden units with weights boldsymbol wi1≤ i ≤ r and output y∈ mathbb R, i.e., y=sumi=1r σ( boldsymbol wimathsf Tboldsymbol x), with activation functions given by low-degree polynomials. In particular, if σ(x) = a0+a1x+a3x3, we prove that no polynomial-time learning algorithm can outperform the trivial predictor that assigns to each example the response variable mathbb E(y), when d3/2ll rll d2. Our conclusion holds for a `natural data distribution', namely standard Gaussian feature vectors boldsymbol x, and output distributed according to a two-layer neural network with random isotropic weights, and under a certain complexity-theoretic assumption on tensor decomposition. Roughly speaking, we assume that no polynomial-time algorithm can substantially outperform current methods for tensor decomposition based on the sum-of-squares hierarchy. We also prove generalizations of this statement for higher degree polynomial activations, and non-random weight vectors. Remarkably, several existing algorithms for learning two-layer networks with rigorous guarantees are based on tensor decomposition. Our results support the idea that this is indeed the core computational difficulty in learning such networks, under the stated generative model for the data. As a side result, we show that under this model learning the network requires accurate learning of its weights, a property that does not hold in a more general setting.