Searcharxiv⌕ Search

arXiv · 2611.22198

Training Neural Networks to be Locally Adaptive

Abstract

Deep neural networks fitted by global empirical risk minimization commit to a single model for the entire covariate space. In a regression problem, if the regression function $f_0$ is inhomogeneous --- varying rapidly in one region, while being nearly flat in another --- no single model can be appropriate everywhere, and the fit degrades where the function varies most rapidly. Here we study an alternative in which, instead of fitting the network globally, we fit it locally. For each query point $x$, a \emph{local neural network} is trained on the $k$ nearest-neighbours of $x$ alone, making the training sample and the fitted function query-dependent. We analyze the performance of the resulting estimator through a localization approach, establishing that the pointwise risk of the estimator with an appropriate choice of $k$ attains the rate $n^{-2α/(2α+d^{*})}$, up to logarithmic factors, when $f_0$ is locally $α$-Hölder with $α\in(0,1]$ and the covariates are supported on a $d^{*}$-dimensional manifold embedded in $\mathbb{R}^{d}$. A matching Le Cam lower bound shows that this rate is pointwise minimax optimal. However, the local rate-optimality requires $k$ to be calibrated to the smoothness of $f_0$, which is unknown in practice. We therefore provide a locally adaptive choice of $k$ that makes the neural network adapt to the local Hölder exponent of $f_0$ and to the intrinsic dimension simultaneously, both of which are potentially varying over the covariate domain. An extensive simulation study demonstrates the dominance of the locally adaptive neural network over a globally trained network and several state-of-the-art methods.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Bilol Banerjee, Soham Sarkar. 2026-11-17. Training Neural Networks to be Locally Adaptive. https://searcharxiv.com/abs/2611.22198

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

On Minimal Observation Time for Recovering Latent Hawkes Networks

Dynamics of interacting systems in engineering, society, and nature often evolve over latent networks that govern which entities can interact. We study the problem of inferring these networks from event-based observations, which arise naturally in finance, seismology, and neuroscience. While there is substantial algorithmic work addressing this important problem, theoretical results are scarce. In this paper we ask the following fundamental question: what is the minimum time that one must observe the dynamics in order to exactly recover the underlying network, as a function of the number $d$ of interacting entities? For a class of stationary Hawkes processes with sparse, weak interactions, we prove that an observation time of order $\log d$ is sufficient and necessary. For the upper bound we construct a two-stage estimator that uses clipped and binned event data for screening, followed by a least-squares refinement, and apply concentration bounds derived from the Poisson cluster representation. For the lower bound we combine Fano's inequality with Jacod's Girsanov formula for point processes on a suitable subclass of networks.

math.ST↗

Bidirectional Random Projections

This paper analyzes bidirectional random projections for ordinary least squares (OLS) regression under the fixed design setting. Let $(X, Y) \in \mathbb{R}^{n\times p}\times\mathbb{R}^{n}$ be a sample and $W \in \mathbb{R}^{n_1\times n}$, $R \in \mathbb{R}^{p\times p_1}$ be two random projections. We develop an expected excess loss bound for the OLS estimator built on $(WXR, WY)$, extending an established bound for the OLS estimator built on $(XR, Y)$. Our analysis shows that our bound is larger and the gap between the two bounds decays at a rate of approximately $O(1/n_1)$. These findings are confirmed by numerical results.

math.ST↗

Asymptotics for Model Selection in Probabilistic Principal Component Analysis

The probabilistic formulation of principal component analysis promises statistically grounded solutions to the problem of selecting the number of principal components. However, developing tractable model selection methods is complicated by the fact that the probabilistic principal component analysis (PPCA) model exhibits non-standard large sample asymptotics at model singularities, where the Fisher information matrix does not have full rank. In this work, tools from singular learning theory are used to provide a complete description of the marginal likelihood asymptotics for PPCA. The singular Bayesian information criterion (sBIC) along with our asymptotic results provides an effective procedure for selecting the number of principal components. In particular, the sBIC corrects for the overpenalization of the standard dimension-based BIC, which leads to too few principal components being selected, while also selecting the smallest true model with probability converging to one. Our framework extends beyond PPCA, where we also provide expressions for the sBIC that can be used to select between PPCA and factor analysis models. In both simulations and on real data the effectiveness of the proposed sBIC methodology is demonstrated.

math.ST↗