Feature Learning in Wide Neural Networks under $μ$P: Identifiability and Sparse-Dictionary Decomposition of the Mean-Field Limit

• URL: http://arxiv.org/abs/2605.24710v1
• Date: Sat, 23 May 2026 19:26:25 GMT
• Title: Feature Learning in Wide Neural Networks under $μ$P: Identifiability and Sparse-Dictionary Decomposition of the Mean-Field Limit
• Abstract summary: We prove global existence and uniqueness of the mean-field limit of noisy gradient descent under $$P.<n>We characterize identifiability of the mean-field limit.<n>We derive the total feature-learning-error decomposition into statistical, optimization, propagation-of-chaos, and sparse-residual components.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We establish four structural results for feature learning in wide two-layer neural networks under the Maximal Update Parametrization ($μ$P). First, we prove global existence and uniqueness of the mean-field limit of noisy gradient descent under $μ$P, identifying the maximal admissible weight $w^*$ on the moment sequence of the initialization as the reciprocal parameter-moment-growth boundary, and hence the largest weighted moment class propagated by the flow. The finite-particle approximation has uniform-in-time squared-Wasserstein rate $O(N^{-1})$. Second, we characterize identifiability of the mean-field limit: two admissible parameter measures induce the same network function in $L^2$ exactly when their active components agree modulo the finite-rank realization symmetry of the architecture. The orbit depth $D^*_{\mathrm{orb}}$ is separated from the moment-variety depth $D^*_{\mathrm{var}}$. Third, under the Barron-Hermite target condition the active support of the long-time limit measure admits a sparse-dictionary decomposition: it is supported on at most $S^*$ atoms modulo finite-rank realization symmetry, with $S^*$ bounded by an explicit coefficient-threshold number. Fourth, we derive the total feature-learning-error decomposition into statistical, optimization, propagation-of-chaos, and sparse-residual components, with a target-dependent Hermite/Barron tail replacing any initialization-only residual. The four results are tied together by an architectural identity: the triple $(w^*, D^*_{\mathrm{orb}}, S^*)$ -- the maximal admissible weight, the orbit identifiability depth, and the sparse-dictionary depth at which the target is realizable -- is the natural learning cell of the architecture-data pair $(σ, ρ)$. The proofs are self-contained except for standard results from $μ$P and mean-field Langevin theory.

Related papers

• A Function-Space Dichotomy for Compositional Learning: Exponential Sub-Optimality of the Neural Tangent Kernel [0.0]
  A persistent empirical observation is that trained neural networks outperform their neural kernel (NTK) limit on tasks with compositional structure.<n>We show a dichotomy between two complexity measures of the target: its $textbfFourier complexity$, which controls NTK kernel regression, and its $textbf complexity$, which controls learning over depth-$L$, width-$w$ ReLU networks with the variation norm of the weights bounded by $R$.
  arXiv  Detail & Related papers  (2026-07-07T15:21:20Z)
• Hardness of High-Dimensional Linear Classification [58.29089693778071]
  We establish new exponential in dimension lower bounds for the Maximum Halfspace Discrepancy problem.<n>Both are fundamental problems in computational geometry and machine learning in their exact and approximate forms.
  arXiv  Detail & Related papers  (2026-03-19T15:53:41Z)
• Quantitative Convergence of Wasserstein Gradient Flows of Kernel Mean Discrepancies [10.511277414974613]
  We study the quantitative convergence of Wasserstein gradient flows of Kernel Mean Discrepancy functionals.<n>Our setting covers in particular the training dynamics of shallow neural networks in the infinite-width and continuous time limit.
  arXiv  Detail & Related papers  (2026-03-02T15:32:54Z)
• An Information-Minimal Geometry for Qubit-Efficient Optimization [0.0]
  We recast qubit-efficient optimization as a geometry problem.<n>Local-consistency problem coincides exactly with the Sherali-Adams level-2 polytope $mathrmSA(2)$.
  arXiv  Detail & Related papers  (2025-11-11T15:38:57Z)
• Arithmetic-Mean $μ$P for Modern Architectures: A Unified Learning-Rate Scale for CNNs and ResNets [9.94514344279733]
  Arithmetic-Mean $mu$P constrains not each individual layer but the network-wide average one-step pre-activation second moment to a constant scale.<n>We prove that, for one- and two-dimensional convolutional networks, the maximal-update learning rate satisfies $etastar(L)propto L-3/2$; with zero padding, boundary effects are constant-level as $Ngg k$.
  arXiv  Detail & Related papers  (2025-10-05T19:22:50Z)
• Emergence and scaling laws in SGD learning of shallow neural networks [64.48316762675141]
  We study the complexity of online gradient descent (SGD) for learning a two-layer neural network with $P$ neurons on isotropic Gaussian data.<n>We provide a precise analysis of SGD dynamics for the training of a student two-layer network to minimize the mean squared error (MSE) objective.
  arXiv  Detail & Related papers  (2025-04-28T16:58:55Z)
• Constructive Universal Approximation and Finite Sample Memorization by Narrow Deep ReLU Networks [0.0]
  We show that any dataset with $N$ distinct points in $mathbbRd$ and $M$ output classes can be exactly classified.<n>We also prove a universal approximation theorem in $Lp(Omega; mathbbRm)$ for any bounded domain.<n>Our results offer a unified and interpretable framework connecting controllability, expressivity, and training dynamics in deep neural networks.
  arXiv  Detail & Related papers  (2024-09-10T14:31:21Z)
• Neural network learns low-dimensional polynomials with SGD near the information-theoretic limit [75.4661041626338]
  We study the problem of gradient descent learning of a single-index target function $f_*(boldsymbolx) = textstylesigma_*left(langleboldsymbolx,boldsymbolthetarangleright)$<n>We prove that a two-layer neural network optimized by an SGD-based algorithm learns $f_*$ with a complexity that is not governed by information exponents.
  arXiv  Detail & Related papers  (2024-06-03T17:56:58Z)
• Learning with Norm Constrained, Over-parameterized, Two-layer Neural Networks [54.177130905659155]
  Recent studies show that a reproducing kernel Hilbert space (RKHS) is not a suitable space to model functions by neural networks. In this paper, we study a suitable function space for over- parameterized two-layer neural networks with bounded norms.
  arXiv  Detail & Related papers  (2024-04-29T15:04:07Z)
• A Mean-Field Analysis of Neural Stochastic Gradient Descent-Ascent for Functional Minimax Optimization [90.87444114491116]
  This paper studies minimax optimization problems defined over infinite-dimensional function classes of overparametricized two-layer neural networks. We address (i) the convergence of the gradient descent-ascent algorithm and (ii) the representation learning of the neural networks. Results show that the feature representation induced by the neural networks is allowed to deviate from the initial one by the magnitude of $O(alpha-1)$, measured in terms of the Wasserstein distance.
  arXiv  Detail & Related papers  (2024-04-18T16:46:08Z)
• Depth Dependence of $\mu$P Learning Rates in ReLU MLPs [72.14317069090407]
  We study the dependence on $n$ and $L$ of the maximal update ($mu$P) learning rate. We find that it has a non-trivial dependence of $L$, scaling like $L-3/2.$
  arXiv  Detail & Related papers  (2023-05-13T01:10:49Z)

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.