Mean-Field Analysis for Learning Subspace-Sparse Polynomials with Gaussian Input
This work addresses theoretical guarantees for neural network learning in high-dimensional settings, but it is incremental as it builds on existing mean-field analysis frameworks.
The paper tackles the problem of learning subspace-sparse polynomials with Gaussian inputs using SGD and two-layer neural networks, establishing a necessary condition for learnability and proving that a slightly stronger condition ensures exponential loss decay to zero.
In this work, we study the mean-field flow for learning subspace-sparse polynomials using stochastic gradient descent and two-layer neural networks, where the input distribution is standard Gaussian and the output only depends on the projection of the input onto a low-dimensional subspace. We establish a necessary condition for SGD-learnability, involving both the characteristics of the target function and the expressiveness of the activation function. In addition, we prove that the condition is almost sufficient, in the sense that a condition slightly stronger than the necessary condition can guarantee the exponential decay of the loss functional to zero.