MLPs at the EOC: Spectrum of the NTK
- URL: http://arxiv.org/abs/2501.13225v1
- Date: Wed, 22 Jan 2025 21:12:51 GMT
- Title: MLPs at the EOC: Spectrum of the NTK
- Authors: Dávid Terjék, Diego González-Sánchez,
- Abstract summary: 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.
- Score: 7.826806223782053
- License:
- Abstract: We study the properties of the Neural Tangent Kernel (NTK) $\overset{\scriptscriptstyle\infty}{K} : \mathbb{R}^{m_0} \times \mathbb{R}^{m_0} \to \mathbb{R}^{m_l \times m_l}$ corresponding to infinitely wide $l$-layer Multilayer Perceptrons (MLPs) taking inputs from $\mathbb{R}^{m_0}$ to outputs in $\mathbb{R}^{m_l}$ equipped with activation functions $\phi(s) = a s + b \vert s \vert$ for some $a,b \in \mathbb{R}$ and initialized at the Edge Of Chaos (EOC). We find that the entries $\overset{\scriptscriptstyle\infty}{K}(x_1,x_2)$ can be approximated by the inverses of the cosine distances of the activations corresponding to $x_1$ and $x_2$ increasingly better as the depth $l$ increases. By quantifying these inverse cosine distances and the spectrum of the matrix containing them, we obtain tight spectral bounds for the NTK matrix $\overset{\scriptscriptstyle\infty}{K} = [\frac{1}{n} \overset{\scriptscriptstyle\infty}{K}(x_{i_1},x_{i_2}) : i_1, i_2 \in [1:n]]$ over a dataset $\{x_1,\cdots,x_n\} \subset \mathbb{R}^{m_0}$, transferred from the inverse cosine distance matrix via our approximation result. Our results show that $\Delta_\phi = \frac{b^2}{a^2+b^2}$ determines the rate at which the condition number of the NTK matrix converges to its limit as depth increases, implying in particular that the absolute value ($\Delta_\phi=1$) is better than the ReLU ($\Delta_\phi=\frac{1}{2}$) in this regard.
Related papers
- MLPs at the EOC: Concentration of the NTK [7.826806223782053]
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.
arXiv Detail & Related papers (2025-01-24T18:58:50Z) - 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) - Optimal Sketching for Residual Error Estimation for Matrix and Vector Norms [50.15964512954274]
We study the problem of residual error estimation for matrix and vector norms using a linear sketch.
We demonstrate that this gives a substantial advantage empirically, for roughly the same sketch size and accuracy as in previous work.
We also show an $Omega(k2/pn1-2/p)$ lower bound for the sparse recovery problem, which is tight up to a $mathrmpoly(log n)$ factor.
arXiv Detail & Related papers (2024-08-16T02:33:07Z) - 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) - A Fast Optimization View: Reformulating Single Layer Attention in LLM
Based on Tensor and SVM Trick, and Solving It in Matrix Multiplication Time [7.613259578185218]
We focus on giving a provable guarantee for the one-layer attention network objective function $L(X,Y).
In a multi-layer LLM network, the matrix $B in mathbbRn times d2$ can be viewed as the output of a layer.
We provide an iterative algorithm to train loss function $L(X,Y)$ up $epsilon$ that runs in $widetildeO( (cal T_mathrmmat(n,d) + d
arXiv Detail & Related papers (2023-09-14T04:23:40Z) - An Over-parameterized Exponential Regression [18.57735939471469]
Recent developments in the field of Large Language Models (LLMs) have sparked interest in the use of exponential activation functions.
We define the neural function $F: mathbbRd times m times mathbbRd times mathbbRd times mathbbRd times mathbbRd times mathbbRd times mathbbRd times mathbbRd
arXiv Detail & Related papers (2023-03-29T07:29:07Z) - 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) - Low-Rank Approximation with $1/\epsilon^{1/3}$ Matrix-Vector Products [58.05771390012827]
We study iterative methods based on Krylov subspaces for low-rank approximation under any Schatten-$p$ norm.
Our main result is an algorithm that uses only $tildeO(k/sqrtepsilon)$ matrix-vector products.
arXiv Detail & Related papers (2022-02-10T16:10:41Z) - 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) - Spectral properties of sample covariance matrices arising from random
matrices with independent non identically distributed columns [50.053491972003656]
It was previously shown that the functionals $texttr(AR(z))$, for $R(z) = (frac1nXXT- zI_p)-1$ and $Ain mathcal M_p$ deterministic, have a standard deviation of order $O(|A|_* / sqrt n)$.
Here, we show that $|mathbb E[R(z)] - tilde R(z)|_F
arXiv Detail & Related papers (2021-09-06T14:21:43Z)
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.