Distance-Based Separability Measure

• Distance-based separability measure is a metric that quantifies how well-separated classes or clusters are by comparing intra- and inter-group distances in a defined metric space.
• It utilizes statistical tools such as the Kolmogorov–Smirnov statistic and divergence metrics to evaluate distribution overlaps, enabling robust cluster validation and complexity analysis.
• Applications span from machine learning tasks like SVM parameter selection and representation learning to quantum information theory using norm-based and entropy measures.

A distance-based separability measure quantifies how “well separated” classes, clusters, or distributions are by analyzing and comparing distances within and between these groups in a vector space or metric space. This concept is foundational across unsupervised learning (cluster validation), supervised classification (dataset complexity analysis), information theory, and quantum information theory. The principal goal is to develop rigorous, interpretable, and computationally tractable statistics that directly evaluate the geometric or probabilistic overlap between classes, clusters, or more general sets in high-dimensional space.

1. Foundational Definitions and Classes of Distance-Based Measures

At its core, a distance-based separability measure compares the distribution of within-group (‘intra-class’ or ‘intra-cluster’) distances to the distribution of between-group (‘inter-class’ or ‘inter-cluster’) distances. Explicitly, for a dataset partitioned into classes $C_1,\dots,C_n$ (where $n\geq2$), the intra-class distance set for class $C_i$ is $\{ \|x-x'\|_2 : x,x'\in C_i, x\neq x' \}$, and the between-class set is $\{ \|x-y\|_2 : x\in C_i, y\notin C_i \}$. Central measures include the Kolmogorov–Smirnov distance between the empirical CDFs of these samples, or, more generally, divergences or metrics between these distance distributions (Guan et al., 2021, Guan et al., 2020, Guan et al., 2021).

In quantum information, the distance to a convex, physically defined set such as the absolutely separable (AS) states serves as the basis for distance-based non-absolute separability (NAS) measures, using functionals such as relative entropy, Bures distance, Hilbert–Schmidt norm, or trace norm (Patra et al., 2022).

Extensions encompass cluster validation, where cluster separability is assessed analogously by distance distributions (Guan et al., 2020, Guan et al., 2021), and domain shift or representation learning, where separability between group encodings is rigorously measured by variance-decomposition ratios (e.g., Cross-Fusion Distance) (Zhang et al., 29 Jan 2026).

2. Formalism: Distance-Based Separability Index (DSI) and Related Measures

The Distance-based Separability Index (DSI) is a prominent archetype rigorously developed for both supervised and unsupervised scenarios (Guan et al., 2021, Guan et al., 2020, Guan et al., 2020, Guan et al., 2021). For a dataset with $n$ classes $C_1,\dots,C_n$ and Euclidean metric, DSI is defined as:

$$\mathrm{DSI}(D) = \frac{1}{n} \sum_{i=1}^n \sup_{t\in\mathbb{R}} |F_{C_i}(t) - G_{C_i}(t)|$$

where $F_{C_i}(t)$ and $G_{C_i}(t)$ are empirical CDFs of intra-class and between-class distances, respectively. DSI takes values in $n\geq2$0: 0 indicates maximal class overlap (distance distributions coincide), and 1 indicates perfect separation (distributions are disjoint).

Variations include:

• Model-weight matrix separability (Frobenius norm between $n\geq2$1 and identity) for neural networks (Yu et al., 2019),
• Cross-Fusion Distance (CFD): a scale-invariant, variance-ratio log metric quantifying separability of latent representations (Zhang et al., 29 Jan 2026),
• Separability and Scatteredness Ratio for SVM parameter selection: signal-to-noise-style log-metric based on class center distance divided by pooled class standard deviation (Shamsi et al., 2023),
• Projection Separability Indices (PSI-P, PSI-ROC, PSI-PR): Mann–Whitney p-value, ROC-AUC, and PR-AUC for one-dimensional class projections (Acevedo et al., 2019).

In quantum resource theory, the NAS measure is $n\geq2$2 for a bona fide matrix distance $n\geq2$3 (e.g., relative entropy, trace norm) (Patra et al., 2022).

3. Mathematical and Statistical Properties

Distance-based measures exhibit several desirable mathematical properties:

• Invariance: DSI and related indices are invariant under translations and uniform scaling (since distances scale multiplicatively), ensuring physical and geometric interpretability (Guan et al., 2021, Zhang et al., 29 Jan 2026).
• Distribution Sensitivity: DSI and variants capture not just means, but full distributional separation, including shape, scale, and higher-order analogues. CFD distinguishes centroid displacement (fusion-altering) from internal dispersion (fusion-preserving) (Zhang et al., 29 Jan 2026).
• Faithfulness: Value 0 if and only if classes or clusters are statistically indistinct (identical distributions of intra- and inter-group distances), and monotonic increase with increasing separation (Guan et al., 2021).
• Robustness: Some measures are insensitive to global scaling, outlier-resistant (given dominance of main mass), and less sensitive to spurious clusters or classes (Zhang et al., 29 Jan 2026).
• Complexity: Naive DSI computation is quadratic in sample size, but can be reduced by subsampling or GPU batching (Guan et al., 2021). CFD is linear in the number of vectors and their dimension (Zhang et al., 29 Jan 2026).
• Monotonicity and Convexity: Distance-based quantum resource measures such as NAS are convex, monotonic under free operations, and invariant under local unitaries (Patra et al., 2022).

4. Practical Instantiations and Algorithmic Workflow

Classical Setting

For a dataset $n\geq2$4 with class labels:

1. Partition data into classes $n\geq2$5.
2. For each $n\geq2$6, compute all intra-class and between-class distances.
3. Compute empirical CDFs $n\geq2$7 (intra) and $n\geq2$8 (inter).
4. Compute the KS statistic $n\geq2$9.
5. Aggregate to obtain the final separability score (mean or min/max over $C_i$0).

For cluster validation, DSI is analogously computed over unsupervised cluster assignments (Guan et al., 2020, Guan et al., 2021).

Deep Learning and Representation Spaces

• For neural network weight matrices, compute Frobenius norm $C_i$1 between $C_i$2 and $C_i$3, directly quantifying angular/norm orthogonality/separability of class-specific weights (Yu et al., 2019).
• In representation learning, Cross-Fusion Distance is computed as $C_i$4 over all representations from two groups (Zhang et al., 29 Jan 2026).

Quantum Information

• Non-absolute separability is assessed as $C_i$5, with $C_i$6 the convex set of absolutely separable states, and $C_i$7 a contractive matrix norm or divergence (relative entropy, Bures, Hilbert–Schmidt, trace norm). Analytic computation uses knowledge of extremal points of $C_i$8 in low-dimensional cases (Patra et al., 2022).

5. Applications and Performance in Practice

Distance-based separability measures are applied in:

• Internal cluster validity: DSI enables objective evaluation and comparison of clusterings without reference to external ground truth, often outperforming or complementing classical indices such as Dunn, Silhouette, or Davies–Bouldin, especially for well-separated or complex-shaped clusters (Guan et al., 2021, Guan et al., 2020).
• Supervised complexity analysis: DSI and PSI-based indices provide classifier-agnostic insight into dataset hardness, robustly ranking toy and real-world datasets by class overlap (Guan et al., 2020, Acevedo et al., 2019).
• Parameter selection: The S&S ratio automates SVM regularization and kernel selection, eliminating expensive cross-validation (Shamsi et al., 2023).
• Representation learning: CFD identifies domain shift and batch effects, correlating more closely with generalization degradation than Wasserstein, MMD, or Hausdorff distances (Zhang et al., 29 Jan 2026).
• Quantum resource theory: Distance-based NAS upper bounds the true entanglement and provides analytic values for symmetric classes such as Werner states (Patra et al., 2022).
• Generative adversarial assessment: DSI, as a distributional overlap metric, can assess GAN performance by comparing generated and real data distributions (Guan et al., 2021).

Empirical studies consistently find that distance-based metrics are effective in tracking class overlap, clustering performance, and domain separation, and in capturing transitions between regimes (as noise or overlap increases, separability measures decrease monotonically with classifier performance) (Guan et al., 2021, Guan et al., 2020, Zhang et al., 29 Jan 2026).

6. Extensions, Limitations, and Open Directions

Extensions

• Metric distribution choices: DSI can be extended to arbitrary metrics (e.g., $C_i$9, $\{ \|x-x'\|_2 : x,x'\in C_i, x\neq x' \}$0, Wasserstein); CFD formalism generalizes to more than two groups and possible hierarchical aggregation (Guan et al., 2021, Zhang et al., 29 Jan 2026).
• Projection-based and ROC/PR analysis: PSIs provide interpretable projections onto lines between class centroids, producing bounded ROC/PR-based scores particularly suited for low-dimensional embeddings (Acevedo et al., 2019).

Limitations

• Quadratic complexity for large datasets: Full pairwise distance computation in DSI incurs $\{ \|x-x'\|_2 : x,x'\in C_i, x\neq x' \}$1 costs, necessitating subsampling or approximate histograms for scalability (Guan et al., 2021, Guan et al., 2020).
• Curse of dimensionality: Distance distributions become concentrated in high $\{ \|x-x'\|_2 : x,x'\in C_i, x\neq x' \}$2, potentially reducing interpretive value unless metrics or samplings are tuned (Guan et al., 2021).
• Ambiguity in complex class topologies: DSI and many distance-based metrics may fail on non-linearly separable (e.g., “circles,” “moons”) or overlapping classes unless combined with other feature-based or topology-aware indices (Xue et al., 2022).
• Over-sensitivity to global shifts: Some indices, especially Hausdorff/Chamfer, are sensitive to global scaling or outliers, whereas CFD and DSI are more robust (Zhang et al., 29 Jan 2026).

Open Directions

• Weighted-aggregation variants of DSI to account for class or cluster imbalance.
• Analytical study of sensitivity to noise and outlier robustness in high-dimensional regimes.
• Connections to independence testing in metric spaces—distance covariance and related functionals extend the scope of separability measures to nonparametric testing of independence between random elements (Jakobsen, 2017).
• Exploration of alternative distributional divergence measures (e.g., Wasserstein, Jensen–Shannon) for deeper geometric and probabilistic insight.

7. Comparative Table of Core Distance-Based Measures

Measure (Acronym)	Mathematical Principle	Primary Application
DSI	KS distance between intra/inter	Clustering, dataset complexity
CFD	Log-variance-ratio of dispersions	Domain shift, embedded fusion
NAS (quantum)	Min. matrix distance to AS set	Quantum resource theory
S&S Ratio	Log(signal-to-dispersion)	SVM parameter selection
PSI-ROC/PR/P	Projection AUC/Mann–Whitney	DR assessment, embedding validation
Weight matrix $\{ \|x-x'\|_2 : x,x'\in C_i, x\neq x' \}$3	Frobenius norm from orthogonality	Neural network separability

Each metric addresses distinct notions of separability and is optimized for interpretability, computational efficiency, or sensitivity to specific geometric or probabilistic structures.

---

References:

• (Patra et al., 2022) (Quantum non-absolute separability)
• (Guan et al., 2021, Guan et al., 2020, Guan et al., 2021, Guan et al., 2020) (DSI theory and applications)
• (Yu et al., 2019) (Weight matrix separability in deep nets)
• (Zhang et al., 29 Jan 2026) (Cross-Fusion Distance in representation learning)
• (Shamsi et al., 2023) (Separability and Scatteredness Ratio for SVM)
• (Acevedo et al., 2019) (Projection-based separability for DR evaluation)
• (Xue et al., 2022) (Critical discussion of distance-based separability measures)
• (Jakobsen, 2017) (Distance covariance, independence, and separability in metric spaces)