Multi-Hypothesis GLRT Framework

• The paper introduces a unified multi-hypothesis GLRT framework that integrates adaptive detection with model order selection via explicit penalty functions.
• It leverages penalties like BIC and GIC to mitigate overfitting and achieve constant false alarm rate through invariance to nuisance parameters.
• Applications in radar detection demonstrate the framework’s robust performance and simplified one-stage implementation compared to traditional methods.

A multi-hypothesis Generalized Likelihood Ratio Test (GLRT) framework provides a unified approach to adaptive detection and classification where decision-making must occur among several possible alternatives, each possibly parameterized by distinct or composite unknowns. These architectures extend the Neyman–Pearson and likelihood ratio paradigms to scenarios involving multiple alternatives, unknown parameters, and model-order uncertainty, offering principled ways to balance detection accuracy with robustness, overfitting control, and invariance to nuisance parameters. This article details the principles, mathematical formulation, penalty constructions, and applications of the multi-hypothesis GLRT framework, referencing both the Kullback–Leibler–motivated approach for radar detection (Addabbo et al., 2020) and the adversarially robust extension for composite hypotheses (Puranik et al., 2021).

1. Formulation of Multi-Hypothesis GLRT

In the canonical multi-hypothesis detection setting, observed data $Z=[z_1,\dots, z_K]\in \mathbb{C}^{N\times K}$ consists of $K$ independent identically distributed measurements. Decisions are drawn among one null hypothesis $H_0$ and $M$ alternative hypotheses $\{H_{1,m}\}_{m=1}^{M}$, where each model specifies a probability density: $$H_0: z_k \sim g_0(z; \theta_0), \quad H_{1,m}: z_k \sim g_{1,m}(z;\theta_{1,m})$$ Under the assumption that all parameters and model orders are known, the optimal test maximizes the average log-likelihood ratio (“compressed LLR”): $$\Lambda_m \triangleq \frac1K \sum_{k=1}^K \log\frac{g_{1,m}(z_k;\theta_{1,m})}{g_0(z_k;\theta_0)} \tag{1}$$ The chosen hypothesis $m^*$ maximizes $\Lambda_m$. The detection threshold $\eta$ is set to achieve a prescribed false alarm rate $K$0 via

$K$1

This structure generalizes directly to settings where the data under each hypothesis are further parameterized or corrupted by unknowns.

2. Penalty-Augmented GLRT and Model Selection

GLRT-based selection without penalty may overfit, especially when the alternative hypotheses differ in parameterization or model order. To prevent systematic over-selection of more complex models, the framework incorporates a penalty term motivated by a Taylor expansion of the cross-entropy part of the Kullback–Leibler divergence. For model $K$2 with parameter dimension $K$3 and nuisance dimension $K$4, the penalized one-stage detector is

$K$5

where

$K$6

with $K$7 the total number of real-valued samples for complex data.

Alternative penalties are defined for different information criteria:

• AIC-type: $K$8
• MDL/NN: $K$9
• GIC$H_0$0: $H_0$1

The BIC and GIC penalties, by growing with both parameter count and $H_0$2, provide effective control of overfitting while maximizing detection probability $H_0$3 (Addabbo et al., 2020).

3. Adaptive Estimation and Decision Rule

When parameters $H_0$4 and $H_0$5 are unknown, maximum likelihood (ML) estimates are substituted per hypothesis: $H_0$6 The compressed LLR is then

$H_0$7

The unified detection rule is

$H_0$8

where $H_0$9, and $M$0 for $M$1 is as above. This decision rule collapses detection, model order selection (MOS), and GLRT into a single stage (Addabbo et al., 2020).

4. Constant False Alarm Rate (CFAR) via Invariance

Under broad regularity conditions, the framework achieves the constant false alarm rate property (CFAR). Specifically, if under $M$2 the nuisance parameters enter only through invariant structures (covariance matrix, subspace, etc.), the tests are functions of maximal invariants whose null distributions do not depend on the unknowns. Therefore, the threshold $M$3 can be set independently of nuisance parameters, ensuring robust control of $M$4 (Addabbo et al., 2020).

5. Applications and Numerical Performance

The framework has been systematically applied to radar detection scenarios involving multiple noise-like jammers, coherent jammers plus targets, and range-spread targets (Addabbo et al., 2020):

Scenario	Statistic/Implementation	Penalty Behavior
Noise-like jammers	Eigenvalue-based closed-form $M$5	BIC/GIC penalties align with TS performance; AIC/NN overfit
Coherent jammers + target	Whitened energy projections (Kelly–Dudgeon)	BIC/GIC one-stage detectors match TS
Range-spread targets	Block-sample covariance determinants	GIC$M$6, tuned $M$7, outperforms TS

A consistent observation is that BIC or GIC$M$8 penalties suppress overfitting and achieve ROC curves similar to or better than two-stage (TS) competitors, while simplifying implementation (no MOS stage needed). For example, in range-spread targets, a single tuned GIC one-stage detector achieves perfect detection ($M$9) at moderate SINR and improves size/position estimation RMSE versus TS (Addabbo et al., 2020).

6. Robust Multi-Hypothesis GLRT under Adversarial Nuisance

The framework generalizes to problems where each hypothesis is subject to an unknown adversarial nuisance, as formalized in a composite $\{H_{1,m}\}_{m=1}^{M}$0-ary hypothesis test: $\{H_{1,m}\}_{m=1}^{M}$1 with $\{H_{1,m}\}_{m=1}^{M}$2 typically an $\{H_{1,m}\}_{m=1}^{M}$3 ball. The test statistic is

$\{H_{1,m}\}_{m=1}^{M}$4

and the GLRT selects

$\{H_{1,m}\}_{m=1}^{M}$5

In Gaussian models with additive $\{H_{1,m}\}_{m=1}^{M}$6-bounded attacks, the “worst-case” attack per hypothesis has a closed-form as a coordinate-wise clipping function; the corresponding minimized cost leads to selecting $\{H_{1,m}\}_{m=1}^{M}$7 by minimizing $\{H_{1,m}\}_{m=1}^{M}$8. Asymptotic analysis shows GLRT’s error probability under the optimal attack matches minimax rates in high dimension. In multi-class cases, the GLRT achieves improved robustness–accuracy tradeoff under both noise-aware and noise-agnostic attacks compared to naive minimum-distance or pairwise-robust linear schemes (Puranik et al., 2021).

7. Unification, Scope, and Practical Implications

The Kullback–Leibler–motivated multi-hypothesis GLRT provides a unified statistical foundation for detection, GLRT-based estimation, and automatic model order selection in complex adaptive scenarios. Its key advantages include:

• Incorporation of parameter dimension via explicit penalties, generalizing LRT, GLRT, and MOS within a single decision rule.
• CFAR properties through invariance principles under broad structural assumptions.
• Empirical effectiveness across radar scenarios, with one-stage implementation outperforming or matching two-stage procedures across $\{H_{1,m}\}_{m=1}^{M}$9 and parameter estimation metrics.
• Robust extension to composite and adversarial settings, with proven minimax or nearly minimax behavior.

A plausible implication is that the framework’s penalty structure, derived via KLD expansion, is likely beneficial in broader model selection and adaptive detection settings well beyond radar, provided the underlying invariance and structural assumptions hold. The simultaneous penalization of parameter redundancy and exploitation of maximal invariance are central to ensuring both overfitting control and operating point robustness (Addabbo et al., 2020, Puranik et al., 2021).