Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Beyond a Single Signal: SPECTREG2, A Unified MultiExpert Anomaly Detector for Unknown Unknowns

Rahul D Ray

Abstract

Epistemic intelligence requires machine learning systems to recognise the limits of their own knowledge and act safely under uncertainty, especially when faced with unknown unknowns. Existing uncertainty quantification methods rely on a single signal such as confidence or density and fail to detect diverse structural anomalies. We introduce SPECTRE-G2, a multi-signal anomaly detector that combines eight complementary signals from a dual-backbone neural network. The architecture includes a spectral normalised Gaussianization encoder, a plain MLP preserving feature geometry, and an ensemble of five models. These produce density, geometry, uncertainty, discriminative, and causal signals. Each signal is normalised using validation statistics and calibrated with synthetic out-of-distribution data. An adaptive top-k fusion selects the most informative signals and averages their scores. Experiments on synthetic, Adult, CIFAR-10, and Gridworld datasets show strong performance across diverse anomaly types, outperforming multiple baselines on AUROC, AUPR, and FPR95. The model is stable across seeds and particularly effective for detecting new variables and confounders. SPECTRE-G2 provides a practical approach for detecting unknown unknowns in open-world settings.

Beyond a Single Signal: SPECTREG2, A Unified MultiExpert Anomaly Detector for Unknown Unknowns

Abstract

Epistemic intelligence requires machine learning systems to recognise the limits of their own knowledge and act safely under uncertainty, especially when faced with unknown unknowns. Existing uncertainty quantification methods rely on a single signal such as confidence or density and fail to detect diverse structural anomalies. We introduce SPECTRE-G2, a multi-signal anomaly detector that combines eight complementary signals from a dual-backbone neural network. The architecture includes a spectral normalised Gaussianization encoder, a plain MLP preserving feature geometry, and an ensemble of five models. These produce density, geometry, uncertainty, discriminative, and causal signals. Each signal is normalised using validation statistics and calibrated with synthetic out-of-distribution data. An adaptive top-k fusion selects the most informative signals and averages their scores. Experiments on synthetic, Adult, CIFAR-10, and Gridworld datasets show strong performance across diverse anomaly types, outperforming multiple baselines on AUROC, AUPR, and FPR95. The model is stable across seeds and particularly effective for detecting new variables and confounders. SPECTRE-G2 provides a practical approach for detecting unknown unknowns in open-world settings.
Paper Structure (71 sections, 17 equations, 3 figures, 3 tables, 1 algorithm)

This paper contains 71 sections, 17 equations, 3 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Surveys and Foundational Reviews
  4. Ensemble and Bayesian Uncertainty Methods
  5. Evidential Deep Learning
  6. Conformal Prediction
  7. Distance‑Based and Causal Methods
  8. Multi‑Signal Fusion and Open‑Set Recognition
  9. Benchmark Datasets and Preprocessing
  10. Synthetic Causal Dataset
  11. Adult Dataset (UCI)
  12. CIFAR‑10 Image Dataset
  13. Gridworld Reinforcement Learning Environment
  14. Data Preprocessing and Splits
  15. Baseline Methods for Uncertainty Quantification
  16. ...and 56 more sections

Figures (3)

  • Figure 1: Architecture of SPECTRE-G2. Input $\mathbf{x}$ is processed by two parallel backbones: a spectral‑normalised GaussEnc (ensemble of 5) and a plain PlainNet. From these we extract eight complementary signals, which are normalised using validation percentiles and corrected for direction with a pseudo‑OOD set. An adaptive top‑$k$ fusion selects the most discriminative signals (top‑$k$ based on validation AUROC) and averages them to produce the final anomaly score $S(\mathbf{x})$. For tabular data, an optional causal signal is also included.
  • Figure 2: Per‑anomaly AUROC of SPECTRE‑G2 compared to 12 baseline methods across four datasets. Higher values (darker green) indicate better performance. SPECTRE‑G2 achieves the highest AUROC on 11 out of 12 anomaly types. The dashed vertical line separates neural network baselines from causal structure learning methods.
  • Figure 3: Ablation study on the Synthetic dataset. Removing any single signal causes only a minor performance drop, confirming the robustness of the multi‑signal fusion. The full model (SPECTRE‑G2) achieves the highest mean AUROC.