mlsanityData to 13 Oct 2026

Test Sample Accuracy Scales with Training Sample Density in Neural Networks

arXiv 2106.08365 ↗First posted 15 Jun 2021cs.LG cs.AI stat.MLcited by 9

Not on the map. Why?

Abstract (arXiv)

Intuitively, one would expect accuracy of a trained neural network's prediction on test samples to correlate with how densely the samples are surrounded by seen training samples in representation space. We find that a bound on empirical training error smoothed across linear activation regions scales inversely with training sample density in representation space. Empirically, we verify this bound is a strong predictor of the inaccuracy of the network's prediction on test samples. For unseen test sets, including those with out-of-distribution samples, ranking test samples by their local region's error bound and discarding samples with the highest bounds raises prediction accuracy by up to 20% in absolute terms for image classification datasets, on average over thresholds.

Key facts. Cited by 9 papers in the map (data through 13 Oct 2026). Builds on 21 of its 32 references that are in the map. First posted on arXiv 15 Jun 2021.

Builds on

21 of 32 references.

References · 21

  1. 2017Neural Discrete Representation Learningcited by 4,680
  2. 2015Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learningcited by 4,430
  3. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,902
  4. 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,371
  5. 2018A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attackscited by 1,475
  6. 2017Enhancing The Reliability of Out-of-distribution Image Detection in Neural Networkscited by 1,327
  7. 2019Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shiftcited by 1,278
  8. 2017Selective Classification for Deep Neural Networkscited by 825
  9. 2018Invariant Information Clustering for Unsupervised Image Classification and Segmentationcited by 455
  10. 2020CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instancescited by 441
  11. 2020Simple and Principled Uncertainty Estimation with Deterministic Deep Learning via Distance Awarenesscited by 389
  12. 2014On the Number of Linear Regions of Deep Neural Networkscited by 380
9 more
  1. 2016On the Expressive Power of Deep Neural Networkscited by 303
  2. 2020Inductive Biases for Deep Learning of Higher-Level Cognitioncited by 273
  3. 2021SSD: A Unified Framework for Self-Supervised Outlier Detectioncited by 264
  4. 2019Residual Flows for Invertible Generative Modelingcited by 257
  5. 2021Invariance Principle Meets Information Bottleneck for Out-of-Distribution Generalizationcited by 220
  6. 2020Classification-Based Anomaly Detection for General Datacited by 210
  7. 2018To Trust Or Not To Trust A Classifiercited by 207
  8. 2019Deep ReLU Networks Have Surprisingly Few Activation Patternscited by 160
  9. 2020Hybrid Models for Open Set Recognitioncited by 109

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2021: 1; 2022: 3; 2023: 2; 2024: 1; 2025: 0; 2026: 2.

9 papers.

  1. 2022Reliable Visual Question Answering: Abstain Rather Than Answer Incorrectlycited by 86
  2. 2022Real-Time Style Modelling of Human Locomotion via Feature-Wise Transformations and Local Motion Phasescited by 72
  3. 2024Deep Networks Always Grok and Here is Whycited by 50
  4. 2023Sensitivity-Aware Amortized Bayesian Inferencecited by 20
  5. 2023RSMT: Real-time Stylized Motion Transition for Characterscited by 19
  6. 2022A Novel Data Augmentation Technique for Out-of-Distribution Sample Detection using Compounded Corruptionscited by 6
  7. 2026Characterizing the Discrete Geometry of ReLU Networkscited by 1
  8. 2021Properties of Minimizing Entropycited by 0
  9. 2026Complexity of Linear Regions in Self-supervised Deep ReLU Networkscited by 0