Basin Separation Score Analysis

• Basin Separation Score is an umbrella term for quantitative metrics that measure the distinct separation between competing basins, reflecting volume, predictability, and boundary complexity.
• The concept is not tied to a single metric but includes proxies such as basin stability, total variation distance, basin entropy, local convexity measures, and repeated-perturbation insulation.
• Researchers leverage these measures to evaluate system robustness, optimization convergence, and dynamic state classification across diverse applications from power grids to quantum eigensolvers.

Taken together, the cited works suggest that “Basin Separation Score” is best understood as an umbrella label for quantitative summaries of how distinctly competing basins of attraction are separated, rather than as a single standardized formal object. In several of the relevant papers, the exact phrase does not appear; instead, the literature formalizes nearby quantities that answer different questions: the probability that a perturbation returns to a target attractor, the Bayes-optimal distinguishability of future-state distributions, the entropy of mixed basin boundaries, the radius of a local basin of convexity, or the probability of avoiding escape under repeated jump perturbations (Schultz et al., 2016, Jones-McCormick, 22 May 2026, Kumar et al., 2020, Kalra et al., 10 Jul 2026, Schultz et al., 2017).

1. Terminological status and family of related quantities

The first point of clarification is negative but decisive: the cited literature repeatedly states that the exact phrase “Basin Separation Score” is not the formal term used in the underlying papers. In "Potentials and Limits to Basin Stability Estimation" (Schultz et al., 2016), the relevant scalar quantity is basin stability, defined as a basin-volume or return-probability measure. In "Detecting Metastable Basins in High Dimensions via Marginal Trajectory Distribution Discrimination" (Jones-McCormick, 22 May 2026), the closest analogue is Bayes-optimal 0-1 classification risk or, equivalently, total variation distance between future-state marginals. In "Unpredictable basin boundaries in restricted six-body problem with square configuration" (Kumar et al., 2020), the closest quantitative proxies are basin entropy $S_b$ and boundary basin entropy $S_{bb}$. In "Characterization of the basin of convexity for multi-snapshot spike deconvolution via variable projection" (Kalra et al., 10 Jul 2026), the relevant object is an explicit basin of convexity with radius $\varrho$. In "Bounding the first exit from the basin: Independence Times and Finite-Time Basin Stability" (Schultz et al., 2017), the key objects are finite-time basin stability and independence time.

Perspective	Closest formal quantity	Representative paper
Basin volume / return probability	$B(A)=\mu(B(A))$, $BS(S)=V_s/V$, node-wise $B_i(K)$	(Schultz et al., 2016, Rakshit et al., 2017, Kim et al., 2016)
Distributional separability	$R^*_{t^*}(x_i,x_j)$, $$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$	(Jones-McCormick, 22 May 2026)
Boundary unpredictability	$S_b$, $S_{bb}$	(Kumar et al., 2020)
Local convexity / optimization basin	$S_{bb}$0, Hessian spectral bounds	(Kalra et al., 10 Jul 2026)
Repeated-perturbation insulation	$S_{bb}$1, $S_{bb}$2	(Schultz et al., 2017)
Basin localization in optimization	Hessian eigenspectrum, $S_{bb}$3, success probability	(Wang, 11 May 2026)

This suggests that the term is not best treated as a single invariant of a landscape. It is instead a family resemblance among several basin-sensitive quantities, each emphasizing a different notion of separation: volume, predictability, statistical distinguishability, local convexity, or robustness under repeated perturbation.

2. Basin volume and return-probability formulations

In the basin-stability literature, the defining idea is probabilistic rather than geometric. For a dynamical system

$S_{bb}$4

with attractor $S_{bb}$5 and basin of attraction $S_{bb}$6, basin stability is

$S_{bb}$7

The interpretation is explicit: basin stability is “the probability that the system will return to $S_{bb}$8” after a perturbation drawn from $S_{bb}$9 (Schultz et al., 2016). The same paper also states that “we are not interested in the basin of attraction’s geometry but only in its volume w.r.t. the measure $\varrho$0”. That sentence sharply distinguishes basin stability from any geometric notion of basin separation.

Because the integral is usually intractable, basin stability is estimated by Monte Carlo sampling. If $\varrho$1 perturbations are drawn and $\varrho$2 trajectories converge to the target attractor, the estimator is

$\varrho$3

The corresponding sampling standard error is

$\varrho$4

The same work emphasizes that sampling error is only one component of total estimation error; approximation, integration, and rounding error can dominate when basin geometry is intricate (Schultz et al., 2016).

A finite-time extension makes the return notion temporal rather than asymptotic. For a return surface $\varrho$5 and a time-tracking Lyapunov construction, finite-time basin stability is

$\varrho$6

The associated independence time

$\varrho$7

quantifies the time after which a perturbed trajectory has probably returned close enough to the attractor that subsequent perturbations can be treated approximately independently. If jump perturbations are at least $\varrho$8 apart, the remain probability is bounded below by

$\varrho$9

In this formulation, separation is operationalized as insulation from basin escape under repeated shocks, rather than as a geometric distance to a boundary (Schultz et al., 2017).

The same basin-volume logic appears in applications. In chimera-state analysis, basin stability is estimated empirically as

$B(A)=\mu(B(A))$0

where $B(A)=\mu(B(A))$1 initial histories converge to a given collective state among $B(A)=\mu(B(A))$2 sampled histories. In small power-grid motifs, node-wise basin stability is treated as a function $B(A)=\mu(B(A))$3 of coupling strength, and transition curves are studied across topologies and producer-consumer placements (Rakshit et al., 2017, Kim et al., 2016).

3. Statistical distinguishability and local convexity

A more direct route to something that behaves like a separation score appears in metastable Markov processes. For a time-homogeneous Markov process with transition kernel

$B(A)=\mu(B(A))$4

the paper on marginal trajectory distribution discrimination reduces basin identification to a two-sample discrimination problem (Jones-McCormick, 22 May 2026). Two initial states $B(A)=\mu(B(A))$5 are compared through their time-$B(A)=\mu(B(A))$6 endpoint marginals

$B(A)=\mu(B(A))$7

The central result is a classification-risk separation theorem: if $B(A)=\mu(B(A))$8 lie in the same basin core, the Bayes-optimal classifier has risk close to $B(A)=\mu(B(A))$9; if they lie in different basin cores, the optimal risk is close to $BS(S)=V_s/V$0. The decision-theoretic and probabilistic forms are equivalent through

$BS(S)=V_s/V$1

Accordingly, the most direct induced pairwise separation score is

$BS(S)=V_s/V$2

Here same-basin behavior corresponds to $BS(S)=V_s/V$3 and $BS(S)=V_s/V$4, whereas different-basin behavior corresponds to $BS(S)=V_s/V$5 and $BS(S)=V_s/V$6. The practical algorithm uses held-out neural classification risk $BS(S)=V_s/V$7 and merges representatives when

$BS(S)=V_s/V$8

This is the most explicit instance in the cited material where basin separation is turned into a quantitative, pairwise discriminability problem (Jones-McCormick, 22 May 2026).

A different but related perspective arises in variable-projection spike deconvolution. There the issue is not pairwise distinguishability of basins but the existence of a certified basin of convexity around the true spike locations. The governing quantity is the minimum separation $BS(S)=V_s/V$9, compared against PSF roughness and amplitude dynamic range via the threshold

$B_i(K)$0

When this holds, the VarProSD objective has an explicit neighborhood

$B_i(K)$1

with radius

$B_i(K)$2

Within this basin, the Hessian satisfies

$B_i(K)$3

and gradient descent converges linearly when initialized inside the neighborhood (Kalra et al., 10 Jul 2026). In this setting, a basin-separation quantity is local, strongly tied to conditioning, and explicitly dependent on sampling bandwidth, PSF smoothness, and spike separation.

4. Boundary complexity, entropy, and geometric pathology

A basin may have large volume and still be poorly separated geometrically. The uncertainty-fraction analysis in basin-stability theory captures this distinction. If initial conditions are known only up to numerical uncertainty $B_i(K)$4, the uncertainty fraction $B_i(K)$5 scales as

$B_i(K)$6

for smooth basin boundaries, but

$B_i(K)$7

for fractal basin boundaries, with $B_i(K)$8 implying final-state sensitivity (Schultz et al., 2016). More extreme pathologies are riddled basins, whose complement intersects every disk in a set of positive measure, and intermingled basins, where any open set intersecting one basin in positive measure also intersects each of the others in positive measure. In such cases, finite-precision basin assignment becomes practically non-deterministic.

The restricted six-body problem gives an explicit entropy-based quantification of poor basin separation. After partitioning the basin image into boxes, the box entropy is

$B_i(K)$9

with total entropy

$R^*_{t^*}(x_i,x_j)$0

where $R^*_{t^*}(x_i,x_j)$1 is the number of boxes containing more than one color (Kumar et al., 2020). The paper uses the criterion that if $R^*_{t^*}(x_i,x_j)$2 or $R^*_{t^*}(x_i,x_j)$3 is greater than $R^*_{t^*}(x_i,x_j)$4, then the basin or its boundaries are fractal. The specific conclusion is that for $R^*_{t^*}(x_i,x_j)$5 and $R^*_{t^*}(x_i,x_j)$6, the basin of attraction is “unpredictable throughout”, while for all values of $R^*_{t^*}(x_i,x_j)$7, $R^*_{t^*}(x_i,x_j)$8, so the basin boundaries are highly unpredictable. The same work also reports Wada boundary evidence, implying that arbitrarily small neighborhoods of boundary points contain multiple basin colors.

A further distinction between volume and geometry appears in the study of mechanically stable particle packings. There, the rank-ordered basin-volume distribution satisfies

$R^*_{t^*}(x_i,x_j)$9

and may extend over $$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$0 decades (Hagh et al., 9 Jul 2026). Yet basin geometry is not exhausted by volume alone. With a Hessian-mode hyper-rectangular compact volume $$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$1, the paper defines a contortion measure

$$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$2

Large variation in $$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$3 shows that equal-volume basins can differ greatly in shape. This suggests that any basin-separation concept based only on volume or accessibility omits potentially decisive information about boundary contortion and local exit directions.

5. Operational proxies in optimization and applied systems

In variational quantum eigensolvers, the language shifts from basin separation to basin localization and basin targeting. The energy landscape is

$$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$4

and basin membership is assessed through the local quadratic expansion

$$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$5

Negative Hessian eigenvalues indicate escape directions toward competing basins; a nonnegative spectrum up to gauge modes indicates a locally convex region associated with the target basin. The paper also uses initialization error

$$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$6

and improvement ratios such as $$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$7–$$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$8 as proxies for whether initialization selected the correlated ground-state basin rather than a competing one (Wang, 11 May 2026). Here the relevant “score” is neither global nor purely geometric; it is a practical diagnostic of correct basin targeting before local refinement.

In multi-band astronomical source separation, the score-matching prior paper explicitly states that it does not define or evaluate an explicit “Basin Separation Score” and does not directly measure distances between optimization basins, attraction-region volumes, or margins between competing source assignments (Sampson et al., 2024). Nevertheless, it offers concrete proxies for basin quality: convergence under severe initialization perturbations, variance of recovered flux and morphology metrics, dependence on blendedness

$$d_{\mathrm{TV}}\!\bigl(P_{t^*}(x_i,\cdot),P_{t^*}(x_j,\cdot)\bigr)$$9

and prior-vs-likelihood dominance through the hallucination score

$S_b$0

These are not separation scores in the strict sense, but they operationalize how easily optimization escapes bad local regions and returns to physically plausible solutions.

In networked dynamical systems, basin-separation ideas often appear through parameterized basin-stability transitions rather than explicit distances. For power-grid swing dynamics, node-wise basin stability $S_b$1 is treated as a function of coupling strength, and the transition curves are embedded in

$S_b$2

The resulting classes correlate more clearly with betweenness and flow betweenness than with degree or edge density (Kim et al., 2016). For chimera states in delay-coupled Mackey-Glass networks, the basin shares of incoherent, chimera, and coherent states are estimated over polynomially parameterized initial history functions, with the collective-state label determined by the strength of incoherence (Rakshit et al., 2017). In both examples, basin separation is read through relative basin shares and transition classes, not through a single geometric margin.

6. Conceptual distinctions, limitations, and synthesis

A recurring misconception is that a basin-based scalar automatically measures geometric basin separation. The literature does not support that identification. Basin stability measures basin volume under a perturbation measure, not boundary smoothness, fractality, Wada structure, riddling, or local distance to a competitor. Two attractors can therefore have identical basin stability and radically different basin geometries (Schultz et al., 2016). Conversely, a high-entropy or high-total-variation boundary diagnostic says little by itself about asymptotic return probability.

Another limitation is context dependence. Distributional separation depends on the chosen observation horizon $S_b$3 and observation space; the metastability paper explicitly notes that same-basin and different-basin distinguishability can reverse if $S_b$4 is chosen outside the ideal within-basin-mixing / pre-transition regime (Jones-McCormick, 22 May 2026). Convexity-basin size depends on PSF descriptors, bandwidth, and amplitude dynamic range (Kalra et al., 10 Jul 2026). Perturbation-based robustness depends on the perturbation distribution $S_b$5, the return surface $S_b$6, and the jump spacing relative to $S_b$7 (Schultz et al., 2017). Numerical reliability can collapse under riddled or intermingled basin geometry, even when Monte Carlo sampling error appears small (Schultz et al., 2016).

Taken together, these works suggest a useful taxonomy of basin-separation quantities. A global return-probability score is represented by $S_b$8, $S_b$9, or $S_{bb}$0. A pairwise separation score is represented by $S_{bb}$1 or $S_{bb}$2. A boundary unpredictability score is represented by $S_{bb}$3 or $S_{bb}$4. A local optimization-basin score is represented by $S_{bb}$5, Hessian spectral positivity, or a curvature bound. An accessibility proxy is represented by initialization error $S_{bb}$6, success probability, convergence robustness, or basin-volume rank distributions (Schultz et al., 2016, Jones-McCormick, 22 May 2026, Kumar et al., 2020, Kalra et al., 10 Jul 2026, Schultz et al., 2017, Wang, 11 May 2026, Hagh et al., 9 Jul 2026). In that sense, Basin Separation Score is best treated not as a single established metric, but as a task-dependent label for whichever of these quantitatively answers the relevant separation question.