A homotopy theorem for incremental stability

Andrey Kharitenko; Fulvio Forni; Rodolphe Sepulchre; Thomas Chaffey

REVIEW 3 major objections 5 minor 2 cited by

A homotopy theorem for incremental stability

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A uniform incremental-gain bound along a one-parameter homotopy proves that a nonlinear feedback loop is incrementally stable on all finite-energy signals.

desk verdict The main homotopy theorem is real and the proof structure holds, but the IQC corollary as printed has a genuine definitional error that invalidates its proof until fixed. read the letter →

arxiv 2412.01580 v1 pith:DWNVVHLD submitted 2024-12-02 math.OC cs.SYeess.SY

Thomas Chaffey , Andrey Kharitenko , Fulvio Forni , Rodolphe Sepulchre This is my paper · ORCID

classification math.OCcs.SYeess.SY MSC 93D2593C1047H10

keywords incrementalstabilityhomotopyscaledrelativegraphintegralquadraticconstraintsgainsmalltheoremfeedbacksystemsL2signalspace

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves a homotopy theorem for incremental stability: given two operators $H_1,H_2\colon L_2\to L_2$ with finite incremental gain, if every partially scaled feedback loop $[H_1,\tau H_2]$ has the same incremental gain bound $\gamma$ for all $\tau\in[0,1]$, then the full feedback system is defined on all of $L_2$ and has incremental gain at most $\gamma$. The result matters because it certifies stability without first assuming well-posedness or causality and without leaving the space of finite-energy signals. The theorem is then used to correct the assumptions of the scaled-relative-graph separation criterion and to derive an incremental version of the classical integral-quadratic-constraint stability theorem. An example shows that the corrected assumptions, strict separation with a positive margin, are genuinely necessary.

What carries the argument

The central object is the interpolated feedback map $[H_1,\tau H_2] = (H_1^{-1}+\tau H_2)^{-1}$ regarded as a relation, scaled by $\tau$ from $0$ to $1$. The proof turns on two pieces: the identity $[H_1,(\tau+\nu)H_2] = [[H_1,\tau H_2],\nu H_2]$, which lets a small increase in feedback be viewed as closing a new small loop around an already stable map, and the incremental small gain theorem, which uses the Banach fixed point theorem to show each small loop is well defined and preserves the incremental gain bound $\gamma$. The scaled relative graph appears in Corollary 1 as the geometric device that turns a strict separation condition into the required uniform bound; its chord property and inverse and sum rules translate distances in the complex plane into output-difference estimates.

What would settle it

The theorem is false if there exist incrementally bounded $H_1,H_2$ satisfying the uniform $\gamma$-bound along all partial feedbacks for every $\tau\in[0,1]$, yet some input in $L_2$ makes the full feedback equations $y=H_1(u-H_2(y))$ have no solution or more than one solution; a search for such a pair, starting from boundary cases like the arctangent nonlinearity in the paper's Example 1, would settle it.

Watch

Extended reading notes

Core claim

The central claim is Theorem 2: given operators $H_1,H_2\colon L_2\to L_2$ with finite incremental gain, if there is a single $\gamma>0$ such that every interpolated feedback $[H_1,\tau H_2]$ has incremental gain at most $\gamma$ on its possibly partial domain for every $\tau\in[0,1]$, then the full feedback $[H_1,H_2]$ is a single-valued map on all of $L_2$ and has incremental gain at most $\gamma$. The two corollaries are that strict separation of the scaled relative graphs of $H_1^{-1}$ and $-\tau H_2$ with a positive margin for every $\tau\in(0,1]$ guarantees incremental stability of the interconnection, and that a bounded LTI forward operator $H_1$ and an incrementally bounded feedback $H_2$ admit an incremental IQC theorem in which a frequency-domain multiplier satisfying the standard inequalities yields bounded incremental gain without any well-posedness or causality assumptions.

Load-bearing premise

The load-bearing premise is that one fixed number $\gamma$ bounds the incremental output difference of every partially scaled feedback loop $[H_1,\tau H_2]$ for all $\tau\in[0,1]$, before those loops are known to be defined on the whole signal space.

Editorial extensions

If this is right

A feedback loop can be certified incrementally stable by checking one uniform inequality along the path from zero feedback to full feedback, without any prior well-posedness or causality proof.
The scaled-relative-graph criterion now requires strict separation with a positive margin along the whole homotopy; the paper's Example 1 shows that separation at a single endpoint is insufficient.
The new incremental IQC theorem applies to nonlinear, incrementally bounded feedback operators with no causality assumptions and recovers the classical IQC theorem in the incremental setting.
Under classical finite-gain, causal, well-posed assumptions, the same homotopy argument still works (Theorem 3), giving a middle ground between incremental and non-incremental analysis.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The proof technique suggests the conclusion would survive if the straight-line scaling $\tau H_2$ were replaced by any continuous path of feedback perturbations, provided a uniform incremental gain bound holds along the path and each small step is a contraction.
Because the SRG corollary gives the explicit bound $1/r_{\min}$ for the gain, strict separation could be used as a computational certificate: discretize the scaled SRGs, check the margin, and obtain a provable incremental stability bound.
The incremental IQC corollary may be directly applicable to neural-network-in-the-loop or other static-nonlinearity systems, where multipliers can be searched computationally and no extended space is needed.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The main result here, Theorem 2, is a genuinely useful incremental homotopy theorem: it gives a clean way to verify incremental stability without extended spaces or causality assumptions, and the bootstrap through Lemma 2 is valid. The SRG separation corollary is also a nice correction of the earlier result, and Example 1 convincingly shows why strict separation is needed. The core proof strategy is sound, and the paper is worth engaging with seriously.

That said, the stress-test note is correct about Corollary 2. The definition of ΔH2(y) includes a τ, and then the vector in condition (5) multiplies that by τ again. So the second block becomes τH2(y1) − τ²H2(y2) rather than τ(H2(y1) − H2(y2)). That means condition (5) does not imply condition (9) of Lemma 7, and the proof's assertion that it does is false. This is not a cosmetic typo; the quadratic form is off. The fix is straightforward—delete the τ inside the definition of ΔH2(y)—but as printed, the incremental IQC theorem is not established. This is the one load-bearing flaw in an otherwise sound paper.

The other issues are minor. Theorem 1's proof has a transposed gain factor: the first inequality should use γ2, not γ1, though the final contraction constant is correct. The induction in Theorem 2 is a bit loose; it should state explicitly that you pick k large enough so that stepping by τ in the allowed interval reaches exactly 1. Corollary 3 is garbled, apparently containing two versions of the same statement. None of these affect the main theorem.

One conceptual point worth noting: condition (ii) of Theorem 2 at τ = 1 already gives the desired incremental gain bound on the domain. The content of the theorem is that the domain is all of L2, which the homotopy argument indeed delivers. That is not a flaw, just a useful thing to keep in mind when reading the statement.

This paper is for people working on nonlinear feedback stability, SRGs, or incremental IQCs. It deserves a serious referee. The central theorem is solid, and the IQC corollary is fixable with a one-line change. I would send it out, expecting a conditional accept after a minor revision.

Referee Report

3 major / 5 minor

Summary. The paper develops an incremental homotopy theorem for feedback interconnections of operators on L2. Theorem 2 states that if two incrementally bounded operators H1, H2 satisfy a uniform incremental gain bound γ along the homotopy path [H1, τH2] for τ ∈ [0,1], then the endpoint feedback [H1,H2] has domain L2 and incremental gain bound γ. The proof bootstraps Theorem 1, an incremental small-gain theorem, using Lemma 2 to compose homotopy steps. Two applications are given: Corollary 1 verifies incremental stability via strict separation of Scaled Relative Graphs, correcting assumptions in [1]; Corollary 2 proposes an incremental IQC stability theorem. A final section relaxes incremental boundedness at the price of well-posedness and causality assumptions.

Significance. The main homotopy theorem is a clean and useful tool: it replaces extended-space and causality assumptions with incremental boundedness, and the SRG separation corollary genuinely corrects two technical assumptions in [1]. The proofs are mostly elementary and do not rely on fitted parameters or circular self-citation; the paper is concise and readable. However, one of the two advertised applications, the incremental IQC theorem (Corollary 2), contains a signal-definition error that invalidates the proof as written. Because that flaw is local and fixable, the underlying contribution remains valuable after revision.

major comments (3)

[Section V, Corollary 2, Eq. (5)] The definition of ΔĤ2(y)(jω) := Ĥ2(y1)(jω) − τĤ2(y2)(jω) is inconsistent with the use of Lemma 7 along the homotopy path [H1, τH2]. For that path, Lemma 7 requires condition (9) with h2 = (Δy, τH2(y1) − τH2(y2)), whose Fourier transform is (Δŷ, τ(Ĥ2(y1) − Ĥ2(y2))). With the definition as written, the second block of the vector in (5) becomes τĤ2(y1) − τ²Ĥ2(y2), which is not the transform of the true incremental feedback signal. Consequently, the claim 'Equation (5) gives condition (9) of Lemma 7' is false as stated. The fix is straightforward—define ΔĤ2(y) := Ĥ2(y1) − Ĥ2(y2)—but until this is made, Corollary 2 is unproven.
[Section III, proof of Theorem 2] The induction step concludes dom([H1,H2]) = L2 from the statement that dom([H1,(ν+kτ)H2]) = L2 for all τ ∈ [0,1/(γγ2)) and positive integers k with ν+kτ ≤ 1. This does not logically imply that the value 1 is reached. The proof needs to explicitly choose, for sufficiently large k, τ := (1−ν)/k, which satisfies τ < 1/(γγ2) whenever k > (1−ν)γγ2, and thereby reach ν + kτ = 1. Without this choice, the displayed induction only covers homotopy parameters strictly below 1.
[Section IV, Lemma 5 and Corollary 1] Lemma 5's statement is internally inconsistent: it assumes ui ∈ dom([H1,H2]) but defines yi via (H1^{-1} + τH2)^{-1}(ui), which is the feedback [H1,τH2]. The intended hypothesis is ui ∈ dom([H1,τH2]), with the separation condition rephrased for τH2. Additionally, the proof of Corollary 1 only states that 1/rmin is an incremental gain bound for [H1,νH2], but Theorem 2 requires condition (ii) for every τ ∈ [0,1]. The same argument actually yields the bound for each [H1,τH2] using strict separation, so the proof should apply Lemma 5 pathwise before invoking Theorem 2.

minor comments (5)

[Section III, Theorem 1 proof] The first displayed inequality in the contraction estimate has the gain factors transposed: it should read ∥H2(H1(x̄)) − H2(H1(x))∥ ≤ γ2∥H1(x̄) − H1(x)∥ ≤ γ1γ2∥x̄ − x∥. The final contraction constant γ1γ2 is correct, but the displayed intermediate step is wrong.
[Section V, Lemma 6 proof] The algebraic identity for σ(y) − σ(x) contains sign errors: the second equality should be ⟨y−x, y⟩ + ⟨x, y−x⟩, not ⟨y−x, y⟩ − ⟨x, y−x⟩, and the subsequent expansion should follow with plus signs. The final bound (7) is correct, but the printed derivation is confusing.
[Section V, Lemma 7 proof] In the final inequality, the term ∥Δy∥ should be squared: −ε(1 + 1/λ²)∥Δy∥². As printed, the dimensions and the subsequent bound are inconsistent.
[Section V, Lemma 7 statement] The conclusion reads 'Then there exists λ > 0 such that...' but λ was already introduced as an incremental gain bound of H1. The wording should be 'Then the following bound holds' or introduce a new symbol for the resulting constant.
[Section VI, Corollary 3] The statement appears to concatenate two different results: the first two sentences claim a strong conclusion from finite gain with zero offset plus strict SRG separation, without the well-posedness and causality assumptions used in Theorem 3; the following 'Suppose (i),(ii),(iii)' then restates a different theorem. This should be split into two separate statements or clearly merged.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Theorem 2's condition (ii) is a genuine homotopy hypothesis, not a fitted or self-referential conclusion; self-citations are contextual only.

full rationale

The derivation chain is not circular. Theorem 2 is a conditional homotopy theorem: condition (ii) hypothesizes a uniform incremental gain bound gamma for every scaled feedback [H1, tau H2] on its unknown domain, and the proof uses the external incremental small-gain theorem and Lemma 2 to extend the domain to all of L2 by induction. The conclusion does contain the tau=1 instance of condition (ii), but the theorem's nontrivial content is the domain-extension statement dom([H1,H2]) = L2; assumption (ii) is a hypothesis to be verified by the corollaries, not a parameter fitted from the conclusion and renamed a prediction. No quantity is fitted to data and then called a prediction. The overlaps with the authors' prior work are contextual: [1] is the result being reproved and corrected, and [8] is mentioned as a related extended-space generalization; neither functions as an unverified load-bearing premise. Corollary 1 rests on separation lemmas credited to [7] and on Lemma 5; Corollary 2 rests on Lemma 6, from external source [14], and Lemma 7. The possible technical inconsistency in Corollary 2 between the tau-scaled definition of Delta H2(y) in condition (5) and the true incremental signal h2 required by Lemma 7 is a correctness concern, not a circularity: it does not amount to assuming the desired conclusion or equating the output with the input by construction. Therefore the paper receives a circularity score of 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper is a pure theory note: no empirical free parameters or fitted constants appear. The central claims rest on standard fixed-point and small-gain results, on SRG composition rules imported from [7], on the standing single-valuedness convention for feedback relations, and on the corollary-specific separation and multiplier hypotheses. No new entities are postulated.

assumptions (5)

domain assumption The negative feedback interconnection [H1,H2] defines a single-valued operator on its domain D subset of L2.
This standing assumption is introduced in Section II before Lemma 1; the theorems extend the domain to all of L2 under their hypotheses, but the relation itself is only defined where the feedback equations have a unique solution.
standard math The Banach fixed point theorem and completeness of L2.
Used in Theorem 1 to prove existence and uniqueness of the fixed point of the contraction map Ku; this is the base of the homotopy induction.
standard math SRG inverse and chord-property sum rules from [7], stated as Lemma 4.
The paper relies on these graphical composition rules to turn SRG separation into incremental gain bounds; property 2 is given a short proof, property 1 is imported from [7].
domain assumption For Corollary 2: H1 is bounded LTI, H2 is incrementally bounded, Pi is Hermitian with L-infinity entries, and the IQC inequalities (5) and (6) hold.
These are the hypotheses of the incremental IQC theorem; they are assumed, not derived, and the proof connects them to the homotopy condition via Lemma 7.
domain assumption Definition 1: strict SRG separation with a uniform positive margin rmin for all tau in (0,1].
This is the corrected assumption replacing non-strict separation in [1]. The paper shows via Example 1 that non-strict separation can lead to infinite incremental gain, so this premise is load-bearing for Corollary 1.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A homotopy theorem for incremental stability." pith.science (2026). https://pith.science/paper/DWNVVHLD

@misc{pith2026241201580,
  author       = {Pith},
  title        = {Pith review of: A homotopy theorem for incremental stability},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DWNVVHLD}},
  note         = {Machine review of arXiv:2412.01580}
}

read the original abstract

A theorem is proved to verify incremental stability of a feedback system via a homotopy from a known incrementally stable system. A first corollary of that result is that incremental stability may be verified by separation of Scaled Relative Graphs, correcting two assumptions in [1, Theorem 2]. A second corollary provides an incremental version of the classical IQC stability theorem.

Figures

Figures reproduced from arXiv: 2412.01580 by the authors.

**Figure 1.** Negative feedback interconnection of H1 and H2. We will make use of the following two technical lemmas. Lemma 1. Given operators H1, H2 : L2 → L2, [H1, H2] = (H −1 1 + H2) −1 . Proof. Applying the definitions of relational inverse and sum, we arrive at (H −1 1 + H2) −1 = {(e + z, y) | y = H1(e), u − e = H2(y)}. Setting u = e+z, we arrive at the definition of [H1, H2]. Lemma 2. Given τ, ν ≥ 0 and operators H1, H2 : L… view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Scaled Relative Graph Analysis of Lur'e Systems and the Generalized Circle Criterion
eess.SY 2024-11 conditional novelty 7.0 of 10

A Nyquist-aware extended Scaled Relative Graph lets SRG analysis handle unstable plants and yields a generalized circle criterion with L2-gain bounds.
Scaled Relative Graph Analysis of General Interconnections of SISO Nonlinear Systems
eess.SY 2025-07 reject novelty 6.0 of 10

The authors introduce an extended Scaled Relative Graph that includes Nyquist encirclement data, enabling stability and L2-gain analysis of feedback interconnections with unstable linear components.

Reference graph

Works this paper leans on

16 extracted references · 9 canonical work pages · cited by 2 Pith papers

[1]

Graphical Nonlinear System Analysis,

T. Chaffey, F. Forni, and R. Sepulchre, “Graphical Nonlinear System Analysis,” IEEE Transactions on Automatic Control , pp. 1–16, 2023. DOI: 10.1109/TAC.2023.3234016

arXiv 2023
[2]

On the input-output stability of time-varying nonlinear feedback systems, part one: Conditions derived using concepts of loop gain, conicity, and positivity,

G. Zames, “On the input-output stability of time-varying nonlinear feedback systems, part one: Conditions derived using concepts of loop gain, conicity, and positivity,” IEEE Transactions on Automatic Control, vol. 11, no. 2, pp. 228–238, 1966. DOI: 10.1109/tac. 1966.1098316

arXiv 1966
[3]

System analysis via integral quadratic constraints,

A. Megretski and A. Rantzer, “System analysis via integral quadratic constraints,” IEEE Transactions on Automatic Control , vol. 42, no. 6, pp. 819–830, 1997. DOI: 10.1109/9.587335

doi:10.1109/9.587335 1997
[4]

System Analysis via Integral Quadratic Constraints Part II,

A. Rantzer and A. Megretski, “System Analysis via Integral Quadratic Constraints Part II,” Lund Institue of Technology, ISRN LUTFD 2 / TFRT– 7559– SE, 1997

work page 1997
[5]

Robustness analysis of nonlinear feedback systems: An input-output approach,

T. Georgiou and M. Smith, “Robustness analysis of nonlinear feedback systems: An input-output approach,” IEEE Transactions on Automatic Control, vol. 42, no. 9, pp. 1200–1221, 1997. DOI: 10 . 1109 / 9 . 623082

work page 1997
[6]

C. A. Desoer and M. Vidyasagar, Feedback Systems: Input–Output Properties. Elsevier, 1975. DOI: 10.1016/b978-0-12-212050- 3.x5001-4

work page doi:10.1016/b978-0-12-212050- 1975
[7]

Scaled relative graphs: Non- expansive operators via 2D Euclidean geometry,

E. K. Ryu, R. Hannah, and W. Yin, “Scaled relative graphs: Non- expansive operators via 2D Euclidean geometry,” Mathematical Pro- gramming, 2021. DOI: 10.1007/s10107-021-01639-w

work page doi:10.1007/s10107-021-01639-w 2021
[8]

On the Scaled Relative Graph Separation for Feedback Incremental Stability,

C. Chen and R. Sepulchre, “On the Scaled Relative Graph Separation for Feedback Incremental Stability,” in Benelux Meeting 2024 , 2024

work page 2024

Show all 16 references

[9]

Linear Matrix Inequalities in Control,

C. Scherer and S. Weiland, “Linear Matrix Inequalities in Control,” 2015

2015
[10]

Robust contraction analysis of nonlinear systems via differential IQC,

R. Wang and I. R. Manchester, “Robust contraction analysis of nonlinear systems via differential IQC,” in 2019 IEEE 58th Conference on Decision and Control (CDC) , Nice, France: IEEE, Dec. 2019, pp. 6766–6771. DOI: 10.1109/CDC40024.2019.9029867

2019
[11]

A semi-infinite optimization problem in harmonic analysis of uncertain systems,

U. T. Jonsson, Chung-Yao Kao, and A. Megretski, “A semi-infinite optimization problem in harmonic analysis of uncertain systems,” in Proceedings of the 2001 American Control Conference , vol. 4, 2001, 3029–3034 vol.4. DOI: 10.1109/ACC.2001.946379

2001
[12]

Integral Quadratic Constraints for Neural Networks,

J. Gronqvist and A. Rantzer, “Integral Quadratic Constraints for Neural Networks,” in 2022 European Control Conference (ECC) , London, United Kingdom: IEEE, Jul. 12, 2022, pp. 1864–1869. DOI: 10 . 23919/ECC55457.2022.9838065

2022
[13]

Kernel-Based Models for System Analysis,

H. J. van Waarde and R. Sepulchre, “Kernel-Based Models for System Analysis,” IEEE Transactions on Automatic Control , vol. 68, no. 9, pp. 5317–5332, Sep. 2023. DOI: 10.1109/TAC.2022.3218944

2023
[14]

On the Role of Well-Posedness in Homotopy Methods for the Stability Analysis of Nonlinear Feedback Systems,

R. A. Freeman, “On the Role of Well-Posedness in Homotopy Methods for the Stability Analysis of Nonlinear Feedback Systems,” in Trends in Nonlinear and Adaptive Control: A Tribute to Laurent Praly for His 65th Birthday , ser. Lecture Notes in Control and Information Sciences, ...

2022 doi
[15]

A link between input-output stability and Lyapunov stability,

V . Fromion, S. Monaco, and D. Normand-Cyrot, “A link between input-output stability and Lyapunov stability,” Systems & Control Letters, vol. 27, no. 4, pp. 243–248, Apr. 15, 1996. DOI: 10.1016/ 0167-6911(95)00046-1

1996
[16]

The Scaled Relative Graph of a Linear Operator

R. Pates. “The Scaled Relative Graph of a Linear Operator.” arXiv: 2106.05650. (2021), preprint

2021 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Graphical Nonlinear System Analysis,

T. Chaffey, F. Forni, and R. Sepulchre, “Graphical Nonlinear System Analysis,” IEEE Transactions on Automatic Control , pp. 1–16, 2023. DOI: 10.1109/TAC.2023.3234016

arXiv 2023

[2] [2]

On the input-output stability of time-varying nonlinear feedback systems, part one: Conditions derived using concepts of loop gain, conicity, and positivity,

G. Zames, “On the input-output stability of time-varying nonlinear feedback systems, part one: Conditions derived using concepts of loop gain, conicity, and positivity,” IEEE Transactions on Automatic Control, vol. 11, no. 2, pp. 228–238, 1966. DOI: 10.1109/tac. 1966.1098316

arXiv 1966

[3] [3]

System analysis via integral quadratic constraints,

A. Megretski and A. Rantzer, “System analysis via integral quadratic constraints,” IEEE Transactions on Automatic Control , vol. 42, no. 6, pp. 819–830, 1997. DOI: 10.1109/9.587335

doi:10.1109/9.587335 1997

[4] [4]

System Analysis via Integral Quadratic Constraints Part II,

A. Rantzer and A. Megretski, “System Analysis via Integral Quadratic Constraints Part II,” Lund Institue of Technology, ISRN LUTFD 2 / TFRT– 7559– SE, 1997

work page 1997

[5] [5]

Robustness analysis of nonlinear feedback systems: An input-output approach,

T. Georgiou and M. Smith, “Robustness analysis of nonlinear feedback systems: An input-output approach,” IEEE Transactions on Automatic Control, vol. 42, no. 9, pp. 1200–1221, 1997. DOI: 10 . 1109 / 9 . 623082

work page 1997

[6] [6]

C. A. Desoer and M. Vidyasagar, Feedback Systems: Input–Output Properties. Elsevier, 1975. DOI: 10.1016/b978-0-12-212050- 3.x5001-4

work page doi:10.1016/b978-0-12-212050- 1975

[7] [7]

Scaled relative graphs: Non- expansive operators via 2D Euclidean geometry,

E. K. Ryu, R. Hannah, and W. Yin, “Scaled relative graphs: Non- expansive operators via 2D Euclidean geometry,” Mathematical Pro- gramming, 2021. DOI: 10.1007/s10107-021-01639-w

work page doi:10.1007/s10107-021-01639-w 2021

[8] [8]

On the Scaled Relative Graph Separation for Feedback Incremental Stability,

C. Chen and R. Sepulchre, “On the Scaled Relative Graph Separation for Feedback Incremental Stability,” in Benelux Meeting 2024 , 2024

work page 2024