Vanishing gradients and overfitting are central problems in machine learning, yet are typically analyzed in asymptotic regimes that obscure their dynamical origins. Here we provide a dynamical description of learning in multi-layer perceptrons (MLPs) via a minimal model inspired by Fukumizu and Amari. We show that training dynamics traverse plateau and near-optimal regions, both organized by saddle structures, before converging to an overfitting regime. Under suitable conditions on the data, this regime collapses to a single attractor modulo symmetry. Furthermore, for finite noisy datasets, convergence to the theoretical optimum is impossible, and the dynamics necessarily settle into an overfitting solution.
Figures & tables
Figure 1: A multi-layer perceptron with two hidden layers of arbitrary size.
Figure 2: A schematic representation of the saddle-saddle-attractor scenario in MLP gradient descent learning. Empirically, the number of positive eigenvalues is smaller near the optimal region than in the plateau region. The overfitting is a stable attractor.
Figure 4: Graphs obtained after training the minimal model for 2 million iterations. In the first column we represent the learning curve; on the right there is a representation of the parameters’ orbits during the training, where red stars are the overfitting points and grey lines model the singular region (dark grey) and optimal region (light gray). Plots in (a) represent the training without observational noise (i.e., τ=0 ), while in (b) we added some small noise with τ=0.2 .
Neural network training dynamics studies how parameters and representations change during optimization and how those changes shape learning, stability, and generalization.
We study the gradient flow training dynamics of two-layer multi-layer perceptrons (MLPs) on anisotropic Gaussian data. Even in this simple setting, solving the dynamics in full is notoriously difficult, so we try something different. We propose a "feature ODE" that might be intuitively expected to describe the learning dynamics of a shallow MLP. We find that MLPs trained to learn simple target functions (low-order Hermite polynomials and staircase functions) actually roughly follow the loss trajectories from this feature ODE. This agreement is a surprise because of the size of the simplification and the fact that it does not come from a rigorous chain of approximations. We discuss this surprise and speculate on its reasons and uses.
In this paper, we study the gradient descent dynamics for jointly training both layers of a one-hidden-layer ReLU network to fit a linear target function. Concretely, we consider a realizable setting where inputs are drawn i.i.d. from a Gaussian distribution and labels follow a planted linear model. This stylized framework captures salient features of end-to-end training in inverse problems and certain auto-encoder models. Despite its apparent simplicity, the dynamics remain poorly understood, in part because the loss landscape contains multiple non-strict saddle points, making it unclear why gradient descent from random initialization reliably escapes bad stationary regions. We provide a detailed characterization of the optimization landscape and prove that gradient descent from a moderately small random initialization-simultaneously training both layers-converges to a global minimizer at a linear rate with order-wise optimal sample complexity. Our analysis tracks the trajectory through three phases: an alignment phase in which hidden weights progressively align with the planted direction while the output weights maintain the correct sign pattern; a growth phase in which the norms of both layers increase while preserving alignment; and a local refinement phase in which the aligned neurons rapidly converge to the planted direction, yielding fast local convergence. To rigorously show that GD avoids non-strict saddles, we develop trajectory-level control arguments for the end-to-end dynamics. In addition, we establish novel uniform concentration results that hold along the entire trajectory, and are essential for obtaining order-wise optimal sample complexity. We corroborate our theory with extensive experiments across a range of configurations.
Berk Tinaz, Changzhi Xie, Mahdi Soltanolkotabi
Department of Electrical and Computer Engineering University of Southern California Los Angeles, CA, USA · Department of Computer Science University of Southern California Los Angeles, CA, USA
The last decade of progress in machine learning (ML), especially the deep learning era, has raised a number of scientific questions that challenge the longstanding dogma of the field. One of the most important riddles was the good empirical generalization of overparameterized models. Overparameterized models are highly complex with respect to the size of the training dataset, which enables them to perfectly fit (i.e., interpolate) even noisy training data. Such interpolation of noisy data is traditionally associated with detrimental overfitting, and yet a wide range of interpolating models -- from simple linear models to deep neural networks -- have been observed to generalize remarkably well on fresh test data. Indeed, the discovery of the double descent phenomenon has revealed that highly overparameterized models can improve over the best underparameterized model in test performance. Understanding learning in this overparameterized regime required new theory and foundational empirical studies, even for the simplest case of the linear model. The underpinnings of this understanding have been laid in foundational analyses of overparameterized linear regression and related statistical learning tasks, mostly published between 2018 and 2022, which resulted in precise analytic characterizations of double descent. This paper provides an overview of the theory of overparameterized ML (henceforth abbreviated as TOPML) by focusing on explaining the most foundational findings through a statistical signal processing perspective. We emphasize the unique aspects that define the TOPML research area as a subfield of modern ML theory and outline interesting open frontiers that remain.
Yehuda Dar, Vidya Muthukumar, Richard G. Baraniuk
Faculty of Computer and Information Science, Ben-Gurion University · Schools of Electrical and Computer Engineering, and Industrial and Systems Engineering, Georgia Institute of Technology · Electrical and Computer Engineering Department, Rice University