MLPs at the EOC: Concentration of the NTK

• URL: http://arxiv.org/abs/2501.14724v1
• Date: Fri, 24 Jan 2025 18:58:50 GMT
• Title: MLPs at the EOC: Concentration of the NTK
• Authors: Dávid Terjék, Diego González-Sánchez,
• Abstract summary: We study the concentration of the Neural Tangent (NTK) $K_theta. We prove that an approximate version of gradient independence holds at finite width. Our results imply that in order to accurately approximate the limit, hidden layer widths have to grow quadratically as $m_k = k2 m$ for some $m in bbN+1$ for sufficient concentration.
• Score: 7.826806223782053
• License:
• Abstract: We study the concentration of the Neural Tangent Kernel (NTK) $K_\theta : \mathbb{R}^{m_0} \times \mathbb{R}^{m_0} \to \mathbb{R}^{m_l \times m_l}$ of $l$-layer Multilayer Perceptrons (MLPs) $N : \mathbb{R}^{m_0} \times \Theta \to \mathbb{R}^{m_l}$ equipped with activation functions $\phi(s) = a s + b \vert s \vert$ for some $a,b \in \mathbb{R}$ with the parameter $\theta \in \Theta$ being initialized at the Edge Of Chaos (EOC). Without relying on the gradient independence assumption that has only been shown to hold asymptotically in the infinitely wide limit, we prove that an approximate version of gradient independence holds at finite width. Showing that the NTK entries $K_\theta(x_{i_1},x_{i_2})$ for $i_1,i_2 \in [1:n]$ over a dataset $\{x_1,\cdots,x_n\} \subset \mathbb{R}^{m_0}$ concentrate simultaneously via maximal inequalities, we prove that the NTK matrix $K(\theta) = [\frac{1}{n} K_\theta(x_{i_1},x_{i_2}) : i_1,i_2 \in [1:n]] \in \mathbb{R}^{nm_l \times nm_l}$ concentrates around its infinitely wide limit $\overset{\scriptscriptstyle\infty}{K} \in \mathbb{R}^{nm_l \times nm_l}$ without the need for linear overparameterization. Our results imply that in order to accurately approximate the limit, hidden layer widths have to grow quadratically as $m_k = k^2 m$ for some $m \in \mathbb{N}+1$ for sufficient concentration. For such MLPs, we obtain the concentration bound $\mathbb{P}( \Vert K(\theta) - \overset{\scriptscriptstyle\infty}{K} \Vert \leq O((\Delta_\phi^{-2} + m_l^{\frac{1}{2}} l) \kappa_\phi^2 m^{-\frac{1}{2}})) \geq 1-O(m^{-1})$ modulo logarithmic terms, where we denoted $\Delta_\phi = \frac{b^2}{a^2+b^2}$ and $\kappa_\phi = \frac{\vert a \vert + \vert b \vert}{\sqrt{a^2 + b^2}}$. This reveals in particular that the absolute value ($\Delta_\phi=1$, $\kappa_\phi=1$) beats the ReLU ($\Delta_\phi=\frac{1}{2}$, $\kappa_\phi=\sqrt{2}$) in terms of the concentration of the NTK.

Related papers

• MLPs at the EOC: Spectrum of the NTK [7.826806223782053]
  We study the properties of the Neuralstyle (NTK) $oversetscriptscriptstyleinftyK. $Delta_phi = fracb2a2+b2$ determines the rate at which the condition number of the NTK matrix converges to its limit as depth increases.
  arXiv  Detail & Related papers  (2025-01-22T21:12:51Z)
• The Communication Complexity of Approximating Matrix Rank [50.6867896228563]
  We show that this problem has randomized communication complexity $Omega(frac1kcdot n2log|mathbbF|)$. As an application, we obtain an $Omega(frac1kcdot n2log|mathbbF|)$ space lower bound for any streaming algorithm with $k$ passes.
  arXiv  Detail & Related papers  (2024-10-26T06:21:42Z)
• In-depth Analysis of Low-rank Matrix Factorisation in a Federated Setting [21.002519159190538]
  We analyze a distributed algorithm to compute a low-rank matrix factorization on $N$ clients. We obtain a global $mathbfV$ in $mathbbRd times r$ common to all clients and a local $mathbfUi$ in $mathbbRn_itimes r$.
  arXiv  Detail & Related papers  (2024-09-13T12:28:42Z)
• Efficient Continual Finite-Sum Minimization [52.5238287567572]
  We propose a key twist into the finite-sum minimization, dubbed as continual finite-sum minimization. Our approach significantly improves upon the $mathcalO(n/epsilon)$ FOs that $mathrmStochasticGradientDescent$ requires. We also prove that there is no natural first-order method with $mathcalOleft(n/epsilonalpharight)$ complexity gradient for $alpha 1/4$, establishing that the first-order complexity of our method is nearly tight.
  arXiv  Detail & Related papers  (2024-06-07T08:26:31Z)
• Dimension Independent Disentanglers from Unentanglement and Applications [55.86191108738564]
  We construct a dimension-independent k-partite disentangler (like) channel from bipartite unentangled input. We show that to capture NEXP, it suffices to have unentangled proofs of the form $| psi rangle = sqrta | sqrt1-a | psi_+ rangle where $| psi_+ rangle has non-negative amplitudes.
  arXiv  Detail & Related papers  (2024-02-23T12:22:03Z)
• Provably learning a multi-head attention layer [55.2904547651831]
  Multi-head attention layer is one of the key components of the transformer architecture that sets it apart from traditional feed-forward models. In this work, we initiate the study of provably learning a multi-head attention layer from random examples. We prove computational lower bounds showing that in the worst case, exponential dependence on $m$ is unavoidable.
  arXiv  Detail & Related papers  (2024-02-06T15:39:09Z)
• Learning a Single Neuron with Adversarial Label Noise via Gradient Descent [50.659479930171585]
  We study a function of the form $mathbfxmapstosigma(mathbfwcdotmathbfx)$ for monotone activations. The goal of the learner is to output a hypothesis vector $mathbfw$ that $F(mathbbw)=C, epsilon$ with high probability.
  arXiv  Detail & Related papers  (2022-06-17T17:55:43Z)
• On Outer Bi-Lipschitz Extensions of Linear Johnson-Lindenstrauss Embeddings of Low-Dimensional Submanifolds of $\mathbb{R}^N$ [0.24366811507669117]
  Let $mathcalM$ be a compact $d$-dimensional submanifold of $mathbbRN$ with reach $tau$ and volume $V_mathcal M$. We prove that a nonlinear function $f: mathbbRN rightarrow mathbbRmm exists with $m leq C left(d / epsilon2right) log left(fracsqrt[d]V_math
  arXiv  Detail & Related papers  (2022-06-07T15:10:46Z)
• Deep Learning in High Dimension: Neural Network Approximation of Analytic Functions in $L^2(\mathbb{R}^d,\gamma_d)$ [0.0]
  We prove expression rates for analytic functions $f:mathbbRdtomathbbR$ in the norm of $L2(mathbbRd,gamma_d)$. We consider in particular ReLU and ReLU$k$ activations for integer $kgeq 2$. As an application, we prove expression rate bounds of deep ReLU-NNs for response surfaces of elliptic PDEs with log-Gaussian random field inputs.
  arXiv  Detail & Related papers  (2021-11-13T09:54:32Z)
• Optimal Mean Estimation without a Variance [103.26777953032537]
  We study the problem of heavy-tailed mean estimation in settings where the variance of the data-generating distribution does not exist. We design an estimator which attains the smallest possible confidence interval as a function of $n,d,delta$.
  arXiv  Detail & Related papers  (2020-11-24T22:39:21Z)
• Bulk-boundary asymptotic equivalence of two strict deformation quantizations [0.0]
  The existence of a strict deformation quantization of $X_k=S(M_k(mathbbC))$ has been proven by both authors and K. Landsman citeLMV. A similar result is known for the symplectic manifold $S2subsetmathbbR3$.
  arXiv  Detail & Related papers  (2020-05-09T12:03:18Z)

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.