Initialization-Dependent Sample Complexity of Linear Predictors and Neural Networks
This work addresses open questions in learning theory for neural networks, providing foundational insights into sample complexity, though it appears incremental in nature.
The paper tackles the sample complexity of vector-valued linear predictors and neural networks, showing that size-independent bounds lead to different behavior than scalar-valued cases and establishing a new convex linear prediction problem learnable without uniform convergence.
We provide several new results on the sample complexity of vector-valued linear predictors (parameterized by a matrix), and more generally neural networks. Focusing on size-independent bounds, where only the Frobenius norm distance of the parameters from some fixed reference matrix $W_0$ is controlled, we show that the sample complexity behavior can be surprisingly different than what we may expect considering the well-studied setting of scalar-valued linear predictors. This also leads to new sample complexity bounds for feed-forward neural networks, tackling some open questions in the literature, and establishing a new convex linear prediction problem that is provably learnable without uniform convergence.