Papers
arxiv:2611.08136

How Far Can Untrained Neural Networks Go in Anomaly Detection? From Tables to Pixels

Published on Nov 9
· Submitted by
Seunghyoung Ryu
on Nov 19
Authors:
,
,
,

Abstract

Untrained neural networks with fixed random nonlinear features and Mahalanobis scoring can detect anomalies competitively on tabular and visual tasks without weight optimization or pretrained representations.

Anomaly detection (AD) is often framed as a representation-learning problem, where strong performance is expected to require learned features or AD-specific optimization. However, this view can overlook the intrinsic separability of anomalies from normal references, even in the absence of learned representations. We therefore investigate how far untrained neural networks can go in anomaly detection. To this end, we propose PRISM, a framework combining fixed random nonlinear features with normal-reference Mahalanobis scoring for tabular detection and dense visual localization. Across 47 tabular datasets in ADBench, PRISM-T achieves the highest mean AUROC (86.4%) and the second-highest mean AUPRC (65.9%) among the compared methods. On MVTec AD and VisA, covering 27 visual categories, PRISM-V achieves mean image-level and pixel-level AUROCs of 90.4% and 96.4%, respectively. Notably, both models achieve these results without neural weight optimization or pretrained representations. A theoretical analysis further characterizes how nonlinear transformations affect the mean and covariance contributions to the Mahalanobis score gap, while empirical analyses examine the resulting score separability. These findings establish PRISM as both a competitive anomaly detector and a prior-to-learning baseline for assessing the added value of learned representations in anomaly detection.

Community

Paper author Paper submitter

Can neural networks detect anomalies without learning?

In this paper, we introduce PRISM, a training-free anomaly detection framework for both tabular (PRISM-T) and visual (PRISM-V) data. Using only a single-layer perceptron for tabular data and a simple CNN for visual data, PRISM combines randomly initialized neural networks with statistical distance-based scoring, without optimizing network weights.
Despite its simplicity, PRISM achieves state-of-the-art mean AUROC across 47 ADBench datasets and competitive performance on MVTec AD and VisA.

These results raise a fundamental question: If untrained neural networks can already achieve such strong performance, how much does feature learning really contribute to anomaly detection?

We hope this work offers a new perspective on the role of learning in anomaly detection and establishes a strong training-free baseline for future research.

Sign up or log in to comment

Get this paper in your agent:

hf papers read 2611.08136
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2611.08136 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2611.08136 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2611.08136 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.