Does Your Neural Network Extrapolate? Feature Engineering as Identifiability Bias for OOD Generalization

• URL: http://arxiv.org/abs/2605.07483v2
• Date: Tue, 12 May 2026 23:48:15 GMT
• Title: Does Your Neural Network Extrapolate? Feature Engineering as Identifiability Bias for OOD Generalization
• Abstract summary: Successful deep neural networks discover salient features of data.<n>We show when and why they fail to learn out-of-distribution (OOD)-relevant representations from an in-distribution (ID) training window.
• Score: 1.0989593035411862
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Successful deep neural networks discover salient features of data. We show when and why they fail to learn out-of-distribution (OOD)-relevant representations from an in-distribution (ID) training window. This requires decoupling feature learning from data-generating-process (DGP) identifiability. From a single training window, OOD extrapolation is non-identifiable: infinitely many DGPs are $\varepsilon$-observationally equivalent on the training data but diverge arbitrarily outside it, and no in-distribution criterion alone reliably breaks the tie. A structural commitment, the feature map, label map, and model class $(\varphi, ψ, \mathcal{M})$, dictates the assumed DGP and governs OOD generalization while leaving ID performance essentially unchanged. When architecture, pretraining, augmentation, input formats, or domain knowledge implicitly inject the missing commitment, the model succeeds. When it cannot infer OOD-relevant structure from ID evidence, it fails. Changing only the representation can make the same architecture, at the same in-distribution loss, differ by ${\sim}520\times$ out of distribution. When the commitment is correct and identifiable, OOD error vanishes. For example, Fourier coordinates turn periodic extrapolation into interpolation on $\mathbb{S}^1$. The same mechanism predicts outcomes in three natural-science settings (mass-action chemistry; Kepler's-third-law exoplanet prediction, $n=2{,}362$; and cross-species coding-DNA detection) and in a 264-run positional-encoding study across Transformer, Mamba, and S4D. Finally, a controlled study shows: correct features are necessary but not sufficient. The model class must express the target, and the transformed training data must cover the relevant representation space.

Related papers

• Epistemic Observability in Language Models [0.0]
  We find that models report highest confidence precisely when they are fabricating.<n>We prove, under formal assumptions, that this is not a capability gap but an observational one.<n>We construct a tensor interface that escapes the impossibility by exporting computational byproducts.
  arXiv  Detail & Related papers  (2026-03-20T21:59:34Z)
• Out-of-Distribution Detection in Molecular Complexes via Diffusion Models for Irregular Graphs [11.928558263824213]
  We present a probabilistic OOD detection framework for complex 3D graph data built on a diffusion model.<n>A single probability-flow ODE produces per-sample log-likelihoods, providing a principled typicality score for distribution shift.<n>We validate the approach on protein-ligand complexes and construct strict OOD datasets.
  arXiv  Detail & Related papers  (2025-12-20T17:56:15Z)
• How Does Unlabeled Data Provably Help Out-of-Distribution Detection? [63.41681272937562]
  Unlabeled in-the-wild data is non-trivial due to the heterogeneity of both in-distribution (ID) and out-of-distribution (OOD) data. This paper introduces a new learning framework SAL (Separate And Learn) that offers both strong theoretical guarantees and empirical effectiveness.
  arXiv  Detail & Related papers  (2024-02-05T20:36:33Z)
• GOODAT: Towards Test-time Graph Out-of-Distribution Detection [103.40396427724667]
  Graph neural networks (GNNs) have found widespread application in modeling graph data across diverse domains. Recent studies have explored graph OOD detection, often focusing on training a specific model or modifying the data on top of a well-trained GNN. This paper introduces a data-centric, unsupervised, and plug-and-play solution that operates independently of training data and modifications of GNN architecture.
  arXiv  Detail & Related papers  (2024-01-10T08:37:39Z)
• Toward Understanding Generative Data Augmentation [16.204251285425478]
  We show that generative data augmentation can enjoy a faster learning rate when the order of divergence term is $o(maxleft( log(m)beta_m, 1 / sqrtm)right)$. We prove that in both cases, though generative data augmentation does not enjoy a faster learning rate, it can improve the learning guarantees at a constant level when the train set is small.
  arXiv  Detail & Related papers  (2023-05-27T13:46:08Z)
• $p$-DkNN: Out-of-Distribution Detection Through Statistical Testing of Deep Representations [32.99800144249333]
  We introduce $p$-DkNN, a novel inference procedure that takes a trained deep neural network and analyzes the similarity structures of its intermediate hidden representations. We find that $p$-DkNN forces adaptive attackers crafting adversarial examples, a form of worst-case OOD inputs, to introduce semantically meaningful changes to the inputs.
  arXiv  Detail & Related papers  (2022-07-25T21:42:08Z)
• Few-Shot Non-Parametric Learning with Deep Latent Variable Model [50.746273235463754]
  We propose Non-Parametric learning by Compression with Latent Variables (NPC-LV) NPC-LV is a learning framework for any dataset with abundant unlabeled data but very few labeled ones. We show that NPC-LV outperforms supervised methods on all three datasets on image classification in low data regime.
  arXiv  Detail & Related papers  (2022-06-23T09:35:03Z)
• CARD: Classification and Regression Diffusion Models [51.0421331214229]
  We introduce classification and regression diffusion (CARD) models, which combine a conditional generative model and a pre-trained conditional mean estimator. We demonstrate the outstanding ability of CARD in conditional distribution prediction with both toy examples and real-world datasets.
  arXiv  Detail & Related papers  (2022-06-15T03:30:38Z)
• Towards IID representation learning and its application on biomedical data [2.91364210254152]
  The widely accepted independent and identically distributed (IID) assumption has been criticized in recent studies on causality. In this paper, we argue that instead of being a questionable assumption, IID is a fundamental task-relevant property that needs to be learned.
  arXiv  Detail & Related papers  (2022-03-01T10:15:14Z)
• Discovering Invariant Rationales for Graph Neural Networks [104.61908788639052]
  Intrinsic interpretability of graph neural networks (GNNs) is to find a small subset of the input graph's features. We propose a new strategy of discovering invariant rationale (DIR) to construct intrinsically interpretable GNNs.
  arXiv  Detail & Related papers  (2022-01-30T16:43:40Z)
• Category-Learning with Context-Augmented Autoencoder [63.05016513788047]
  Finding an interpretable non-redundant representation of real-world data is one of the key problems in Machine Learning. We propose a novel method of using data augmentations when training autoencoders. We train a Variational Autoencoder in such a way, that it makes transformation outcome predictable by auxiliary network.
  arXiv  Detail & Related papers  (2020-10-10T14:04:44Z)

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.