The merged-staircase property: a necessary and nearly sufficient condition for SGD learning of sparse functions on two-layer neural networks

• URL: http://arxiv.org/abs/2202.08658v2
• Date: Mon, 26 Aug 2024 23:24:52 GMT
• Title: The merged-staircase property: a necessary and nearly sufficient condition for SGD learning of sparse functions on two-layer neural networks
• Authors: Emmanuel Abbe, Enric Boix-Adsera, Theodor Misiakiewicz,
• Abstract summary: We study SGD-learnability with $O(d)$ sample complexity in a large ambient dimension. Our main results characterize a hierarchical property, the "merged-staircase property", that is both necessary and nearly sufficient for learning in this setting. Key tools are a new "dimension-free" dynamics approximation that applies to functions defined on a latent low-dimensional subspace.
• Score: 19.899987851661354
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: It is currently known how to characterize functions that neural networks can learn with SGD for two extremal parameterizations: neural networks in the linear regime, and neural networks with no structural constraints. However, for the main parametrization of interest (non-linear but regular networks) no tight characterization has yet been achieved, despite significant developments. We take a step in this direction by considering depth-2 neural networks trained by SGD in the mean-field regime. We consider functions on binary inputs that depend on a latent low-dimensional subspace (i.e., small number of coordinates). This regime is of interest since it is poorly understood how neural networks routinely tackle high-dimensional datasets and adapt to latent low-dimensional structure without suffering from the curse of dimensionality. Accordingly, we study SGD-learnability with $O(d)$ sample complexity in a large ambient dimension $d$. Our main results characterize a hierarchical property, the "merged-staircase property", that is both necessary and nearly sufficient for learning in this setting. We further show that non-linear training is necessary: for this class of functions, linear methods on any feature map (e.g., the NTK) are not capable of learning efficiently. The key tools are a new "dimension-free" dynamics approximation result that applies to functions defined on a latent space of low-dimension, a proof of global convergence based on polynomial identity testing, and an improvement of lower bounds against linear methods for non-almost orthogonal functions.

Related papers

• Visualising Feature Learning in Deep Neural Networks by Diagonalizing the Forward Feature Map [4.776836972093627]
  We present a method for analysing feature learning by decomposing deep neural networks (DNNs) We find that DNNs converge to a minimal feature (MF) regime dominated by a number of eigenfunctions equal to the number of classes. We recast the phenomenon of neural collapse into a kernel picture which can be extended to broader tasks such as regression.
  arXiv  Detail & Related papers  (2024-10-05T18:53:48Z)
• Preconditioned Gradient Descent Finds Over-Parameterized Neural Networks with Sharp Generalization for Nonparametric Regression [8.130817534654089]
  We consider nonparametric regression by a two-layer neural network trained by gradient descent (GD) or its variant in this paper. We show that, if the neural network is trained with a novel Preconditioned Gradient Descent (PGD) with early stopping and the target function has spectral bias widely studied in the deep learning literature, the trained network renders a particularly sharp generalization bound with a minimax optimal rate of $cO(1/n4alpha/(4alpha+1)$.
  arXiv  Detail & Related papers  (2024-07-16T03:38:34Z)
• Nonparametric regression using over-parameterized shallow ReLU neural networks [10.339057554827392]
  We show that neural networks can achieve minimax optimal rates of convergence (up to logarithmic factors) for learning functions from certain smooth function classes. It is assumed that the regression function is from the H"older space with smoothness $alpha(d+3)/2$ or a variation space corresponding to shallow neural networks. As a byproduct, we derive a new size-independent bound for the local Rademacher complexity of shallow ReLU neural networks.
  arXiv  Detail & Related papers  (2023-06-14T07:42:37Z)
• ReLU Neural Networks with Linear Layers are Biased Towards Single- and Multi-Index Models [9.96121040675476]
  This manuscript explores how properties of functions learned by neural networks of depth greater than two layers affect predictions. Our framework considers a family of networks of varying depths that all have the same capacity but different representation costs.
  arXiv  Detail & Related papers  (2023-05-24T22:10:12Z)
• Learning Low Dimensional State Spaces with Overparameterized Recurrent Neural Nets [57.06026574261203]
  We provide theoretical evidence for learning low-dimensional state spaces, which can also model long-term memory. Experiments corroborate our theory, demonstrating extrapolation via learning low-dimensional state spaces with both linear and non-linear RNNs.
  arXiv  Detail & Related papers  (2022-10-25T14:45:15Z)
• Exploring Linear Feature Disentanglement For Neural Networks [63.20827189693117]
  Non-linear activation functions, e.g., Sigmoid, ReLU, and Tanh, have achieved great success in neural networks (NNs) Due to the complex non-linear characteristic of samples, the objective of those activation functions is to project samples from their original feature space to a linear separable feature space. This phenomenon ignites our interest in exploring whether all features need to be transformed by all non-linear functions in current typical NNs.
  arXiv  Detail & Related papers  (2022-03-22T13:09:17Z)
• Deep Networks Provably Classify Data on Curves [12.309532551321334]
  We study a model problem that uses a deep fully-connected neural network to classify data drawn from two disjoint smooth curves on the unit sphere. We prove that when (i) the network depth is large to certain properties that set the difficulty of the problem and (ii) the network width and number of samples is intrinsic in the relative depth, randomly-d gradient descent quickly learns to correctly classify all points on the two curves with high probability.
  arXiv  Detail & Related papers  (2021-07-29T20:40:04Z)
• Going Beyond Linear RL: Sample Efficient Neural Function Approximation [76.57464214864756]
  We study function approximation with two-layer neural networks. Our results significantly improve upon what can be attained with linear (or eluder dimension) methods.
  arXiv  Detail & Related papers  (2021-07-14T03:03:56Z)
• Topological obstructions in neural networks learning [67.8848058842671]
  We study global properties of the loss gradient function flow. We use topological data analysis of the loss function and its Morse complex to relate local behavior along gradient trajectories with global properties of the loss surface.
  arXiv  Detail & Related papers  (2020-12-31T18:53:25Z)
• Modeling from Features: a Mean-field Framework for Over-parameterized Deep Neural Networks [54.27962244835622]
  This paper proposes a new mean-field framework for over- parameterized deep neural networks (DNNs) In this framework, a DNN is represented by probability measures and functions over its features in the continuous limit. We illustrate the framework via the standard DNN and the Residual Network (Res-Net) architectures.
  arXiv  Detail & Related papers  (2020-07-03T01:37:16Z)
• Measuring Model Complexity of Neural Networks with Curve Activation Functions [100.98319505253797]
  We propose the linear approximation neural network (LANN) to approximate a given deep model with curve activation function. We experimentally explore the training process of neural networks and detect overfitting. We find that the $L1$ and $L2$ regularizations suppress the increase of model complexity.
  arXiv  Detail & Related papers  (2020-06-16T07:38:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.