Pith. sign in

REVIEW 5 major objections 6 minor 26 references

Constrained Hamiltonian Systems on Observation-Induced Fiber Bundles: Theory of Symmetry and Integrability

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that partially observed constrained Hamiltonian systems can be modeled as symplectic fiber bundles, carrying Dirac constraints, reduction, integrability, Lax pairs, and Noether symmetries.

desk verdict The central geometric construction is internally dimensionally inconsistent—fibers are k-dimensional while the symplectic structure needs 2k—so the framework collapses before the topological shortcuts are even reached. read the letter →

arxiv 2505.22824 v1 pith:JGYP45ZE submitted 2025-05-28 math.GM

Dongzhe Zheng This is my paper
classification math.GM MSC 53D2037J3570H4553C0570G45
keywords observation-inducedfiberbundleconstrainedHamiltoniansystemsymplecticreductionArnold-LiouvilletheoremLaxpairNoether'sControlBarrierFunctionspartialobservation
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

The paper tries to establish that a constrained Hamiltonian system whose states are only partially observed is best described not by a constraint submanifold with noise added, but by a fiber bundle over the state space whose fibers encode all possible observation values. On this observation-induced fiber bundle, the author claims, Dirac's constraint classification, Marsden-Weinstein reduction, Arnold-Liouville integrability, Lax-pair construction, and Noether's theorem all hold in adapted form. If true, this would give a single geometric language for safety-critical control, filtering, and classical mechanics under incomplete information, turning observation uncertainty from an external disturbance into intrinsic fiber geometry. The author honestly notes that global and stochastic extensions remain open.

What carries the argument

The central object is the observation-induced fiber bundle $\pi: E_W \to W$, where $W$ is a working region of a $2n$-dimensional symplectic manifold and each fiber $\pi^{-1}(x)$ is a ball of radius $\delta(x)$ in the cotangent space of the observation manifold at $h(x)$. The argument is carried by an observation-adaptive Ehresmann connection that is metric-compatible, together with the explicit symplectic form $\omega_E = \pi^*\omega_M + \omega_{\mathrm{fib}} + \Omega_{\mathrm{mix}}$, whose mixing coefficients are curvature pairings of the connection; the structure functions $C, D, E, F$ built from these pairings satisfy symmetry relations that make $\omega_E$ closed. This machinery lets the author translate state constraints and observation constraints into a single Poisson structure on the total space, and then run the classical reduction and integrability arguments on that structure.

What would settle it

Take W to be a compact connected open set with smooth boundary that is not contractible, such as an annular region {x ∈ $R^{4}$ : 1 < |x| < 2} with a rank-2 observation map, and compute the second Stiefel-Whitney class of TE_W on W. If the class is non-zero, Theorem 9's existence condition fails and the claimed symplectic structure cannot be guaranteed; if it is zero, the contractibility premise used in the proof is nonetheless false, so the proof of the existence theorem as written does not cover this case.

Watch

Extended reading notes

Core claim

The central claim is that every partially observed constrained Hamiltonian system carries a symplectic fiber-bundle structure, provided the observation map is a local diffeomorphism on a compact connected working region with smooth boundary. The construction pairs the base symplectic form with a fiber symplectic form and a curvature mixing term, producing a closed 2-form that the author shows is non-degenerate inside a safety subregion. On this structure, constraints are functions on the total space, and the paper derives a Dirac-style first/second-class classification, a Marsden-Weinstein reduction theorem for observation symmetry groups, an Arnold-Liouville theorem with action-angle variables on tori, a Lax pair depending on observation uncertainty parameters, and a Noether theorem producing observation-dependent conserved quantities. The author also argues that classical integrable systems such as the Toda lattice and rigid body dynamics fit the framework, and that modern safety-critical control algorithms inherit the geometric guarantees.

Load-bearing premise

The proof assumes without support that the working region has no holes (is contractible), while the stated assumptions only guarantee it is a connected region with a smooth boundary; an annular region is a counterexample.

Editorial extensions

If this is right

  • Dirac's first/second-class distinction and bracket machinery transfer to the total space, so state constraints and observation constraints can be treated by one Poisson structure.
  • Marsden-Weinstein reduction goes through for observation symmetry groups, yielding reduced phase spaces that are again observation fiber bundles.
  • Complete integrability under partial observation is characterized geometrically; compact joint level sets are tori with action-angle variables, and Lax pairs exist that depend on observation uncertainty and reduce to the classical ones as uncertainty vanishes.
  • Noether's theorem holds in observation-dependent form: each continuous symmetry of the Hamiltonian produces a conserved quantity built from the moment map and canonical 1-form.
  • Control Barrier Functions and model-predictive safety constraints acquire symplectic and bundle-geometric foundations, with boundary degeneration acting as a safety buffer.

Reading between the lines

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

  • If the framework is right, sensor uncertainty is not a nuisance to estimate away but a geometric field; this suggests that symplectic-integrator style control laws should outperform filtering-plus-then-control pipelines in preserving long-term invariants.
  • The Euler-class obstruction in Theorem 28 implies a concrete design rule: sensor configurations whose pullback tangent bundle has non-zero Euler class cannot support a global observation bundle, so observability is a topological, not just a rank, condition.
  • The paper's regularized second-class constraints (exponential decay of constraint violation with rate $\alpha/(\mu^2+\epsilon)$) read like a dissipation term; one could test whether the small-$\alpha$ limit reproduces Dirac's ideal constraint surface or a metriplectic perturbation of it.
  • The Toda-lattice example makes a sharp quantitative prediction — observation error $\epsilon_i$ must satisfy $|\epsilon_i| \le \min_i |p_i-p_{i+1}|/(2\sigma_0)$ for integrability to survive — which is checkable by direct numerical simulation of the noisy lattice.
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.

Referee Report

5 major / 6 minor

Summary. The paper proposes a 'complete geometric theoretical framework' for constrained Hamiltonian systems under partial observation, based on observation-induced fiber bundles E_W -> W whose fibers encode observation uncertainty. It claims to extend Dirac constraint theory, Marsden-Weinstein reduction, Arnold-Liouville integrability, Lax pair theory, and Noether's theorem to this setting, and it presents applications to control and robotics. The central construction is a symplectic structure on E_W built from the base symplectic form, a 'fiber standard symplectic form', and curvature mixing terms, followed by Dirac-type classification of constraints, reduction, integrability, and Lax pair results.

Significance. If the framework were correct, it would indeed unify observation uncertainty with constrained Hamiltonian dynamics in a geometric way, and the claimed applications to safety-critical control would be of broad interest. The paper is also commendable for attempting to make the construction explicit, e.g., by writing down structure functions, connection coefficients, and Lax matrices, and for not relying on empirical fitting. However, the central geometric object is internally dimensionally inconsistent, and several load-bearing proofs contain gaps or errors; as it stands, the framework does not provide a valid symplectic or Poisson structure on the defined bundle. The manuscript is therefore not a sound foundation for the claimed extensions.

major comments (5)
  1. [Definition 1, Definition 8, Theorem 9, Theorem 10] There is a fatal dimension mismatch in the central construction. Definition 1(6) defines the total space as E_W = {(x, ξ) : x ∈ W, ξ ∈ T*_{h(x)}Y, ||ξ|| ≤ δ(x)}. Since T*_{h(x)}Y is k-dimensional, this gives dim E_W = dim W + k = 2n + k. However, Theorem 9(1) asserts dim E_W = 2(n+k), and Definition 8 treats each fiber as a 2k-dimensional cotangent bundle with coordinates (ξ^a, π_a) and canonical symplectic form Σ dξ^a ∧ dπ_a. A single cotangent space T*_{h(x)}Y cannot carry such a 2k-dimensional symplectic form; the 'momentum' coordinates π_a would have to live in T*(T*_{h(x)}Y), not in T*_{h(x)}Y. Consequently, the 2-form ω_E in Theorem 10 is not a well-defined nondegenerate 2-form on the E_W of Definition 1. All subsequent results that use a symplectic structure on E_W — Proposition 14, Theorem 15, Theorem 21, Theorem 22, Theorem 23, Theorem 24 — are therefore unsupported.
  2. [Theorem 9, Step 2 and Theorem 10, Step 1] The proof relies on the claim that 'Due to contractibility of W, second Stiefel-Whitney class w2(TE_W) automatically vanishes.' Assumption 1 only assumes that W is a connected open set with compact closure and smooth boundary; such a set need not be contractible. An open annulus in R^2 is a counterexample. No proof of contractibility is supplied. Thus the vanishing of w2(TE_W) is not established, and the existence of the symplectic structure in Theorem 10, which depends on this step, is not proved.
  3. [Proposition 16, Definition 14, Theorem 19, Proposition 20] The regularized treatment of second-class constraints is internally inconsistent. Proposition 16 correctly computes dΦ/dt = {H,Φ}_E + λ{Φ,Φ}_E = {H,Φ}_E, since {Φ,Φ}_E = 0 by antisymmetry. But Definition 14 defines λ = -({H,Φ}_E + αΦ)/(||∇^E Φ||^2 + ε), so the λ term disappears from dΦ/dt. The dynamics of Φ are therefore independent of the regularization parameters α and ε, and the claimed exponential decay |Φ(t)| ≤ |Φ(0)| exp(-αt/(μ^2+ε)) in Proposition 20 cannot follow from the stated equations of motion. The regularized system is dynamically identical to the unregularized one, so Theorem 19(1) is unsupported.
  4. [Example 1, Step 4 and Example 3, Step 'O(λ^0) term'] The order-by-order verification of the Toda Lax pair contains a sign error. In both examples, the (1,2) element of the zero-curvature equation is computed as LHS = e^{y1+ε1}(p1 - p2 + dε1/dt) and RHS = e^{y1+ε1}(p2 - p1). Equality therefore requires dε1/dt = 2(p2 - p1), not dε1/dt = 2(p1 - p2) as stated. Since this equality is used to determine the 'dynamical evolution law of observation error,' the verification at O(λ^0) fails, and the claimed Lax integrability of the example is not established.
  5. [Lemma 27 and Theorem 28] Lemma 27 states that for an isometric embedding ι : Y -> R^N, the normal bundle satisfies e(N_{ι(Y)}) = 0 because 'the normal bundle is parallelizable.' This is false in general: an isometrically embedded submanifold of R^N can have a nontrivial normal bundle (e.g., embeddings of RP^2 require nonorientable normal bundles). The Euler class of the normal bundle does not have to vanish. Since Theorem 28 and Remark 21 use this reduction to justify treating Y = R^k, the topological classification of observation fiber bundles for general observation manifolds is unsupported.
minor comments (6)
  1. [Section 7.2 and Section 7.4] The text contains unresolved placeholders 'Section ??' twice (in Section 7.2 and Section 7.4), which should be fixed to actual section numbers.
  2. [Definition 2(4) and Definition 5] The notation ∥R^∇∥L∞(W) is used without specifying the norm on the curvature tensor; given the mixture of base, fiber, and mixed components, this should be defined precisely.
  3. [Remark 14] The asserted energy identity dH/dt = -α δ_boundary ||∇H||^2 ≤ 0 is stated without derivation and appears inconsistent with Hamiltonian dynamics on a degenerate symplectic form; a derivation or precise hypothesis is needed.
  4. [Throughout] The symbol δ is used for the observation uncertainty function, for the boundary degeneration parameter δ_boundary, and for Lax matrix perturbation terms δ_i; this overloading makes several formulas ambiguous.
  5. [Example 2] The definition ℓ0 = min_{i,j} |q_i(0) - q_j(0)| would be zero for i = j; it should presumably be min_{i≠j} |q_i(0) - q_j(0)|, and similarly for related quantities.
  6. [Theorem 7, necessity proof] In the necessity proof of condition (C1), the text says that 'appropriate choice of constraint potential' ensures Φ(x_n, ξ_n) < 0; this is not a proof, since the constraint potential Φ is not under the control of the theorem's hypotheses.

Circularity Check

1 steps flagged · score 2.0 of 10

Central geometric derivation is self-contained; the only circular element is the use of the authors' own ICML paper [25] as the source of application validation.

  1. self citation load bearing [Section 1.4 (contribution 3) and Section 6.2 (introductory paragraph)]
    "providing complete mathematical support for ’Learning Dynamics under Environmental Constraints via Measurement-Induced Bundle Structures’[25], validating the theory’s effectiveness in practical systems such as soft robotics, robotic arm control, and quadrotor navigation ... Following are successful cases demonstrated in [25] under guidance of this theoretical framework:"

    The paper presents [25] as the external validation of this theory’s effectiveness, but [25] is by the same first author and the Section 6.2 examples are introduced as “successful cases demonstrated in [25] under guidance of this theoretical framework.” The validation evidence is therefore generated by the framework it is supposed to validate; Section 7.4 completes the loop by saying that the present framework provides the foundation for [25]. No independent simulations or experiments appear in this paper. This makes the “Application Verification” contribution circular, although the main existence, reduction, Noether, and Lax-pair proofs are derived from differential-geometric inputs and do not reduce to [25].

full rationale

The main derivation chain, from Assumption 1 through the connection construction, symplectic-form construction, Poisson brackets, Dirac classification, Marsden-Weinstein-type reduction, Noether theorem, and Lax-pair theorems, does not use [25] and does not fit parameters to observations. The claimed theorems are built on standard differential-geometric ingredients (cotangent fibers, connections, characteristic classes, Poisson manifolds), so the central claim is not circular. The one genuine circularity is the application-validation loop: Section 1.4 cites the authors’ own ICML paper [25] as validating the theory’s effectiveness, Section 6.2 presents that paper’s cases as “under guidance of this theoretical framework,” and Section 7.4 in turn says the present framework provides the foundation for [25]. That mutual citation does not enter the geometric proofs, so the score stays low. Separately, there are internal-consistency and support gaps that are not circularity: Theorem 9’s proof invokes contractibility of W although Assumption 1 only asserts connectedness and compactness of a compact open set, and Definition 1’s fibers are k-dimensional cotangent spaces while Definition 8/Theorem 10 require a 2k-dimensional symplectic fiber structure. Those are correctness concerns, not reductions to the paper’s own inputs, so they do not increase the circularity score.

Assumptions & free parameters 4 free parameters · 6 assumptions · 3 invented entities

The central construction rests on substantial domain assumptions: compact working region, constant-rank observation map, curvature bounds on a specially chosen connection, and monotonicity properties of the constraint potential. In addition, two topological shortcuts (contractibility of W and trivial normal bundle in the Whitney embedding) are not justified. The framework introduces new geometric objects with no independent empirical evidence, and the example constants are chosen by hand rather than fitted to data or derived.

free parameters (4)
  • uncertainty function δ(x), bounds δ_min and Δ
    Chosen by hand; defines fiber radius and boundary safety buffer; central to bundle and properness theorems, with no data-derived value.
  • fiber metric family ρ_x
    Free geometric input on each fiber; determines connection coefficients, mixed symplectic terms, and gradient norm; no independent measurement.
  • boundary cutoff function χ and boundary degeneration parameter
    Introduced ad hoc to interpolate between interior symplectic form and boundary degenerate form; no derivation from physical data.
  • example constants δ0, α, β, κ, w_i = various
    Used in Toda and robot examples to define observation uncertainty, correlation decay, and constraint strength; not derived or fitted to measurements.
assumptions (6)
  • domain assumption Assumption 1: M is a 2n-dimensional symplectic manifold, W is a connected compact open set with smooth boundary, h:W->Y is a constant-rank local diffeomorphism onto its image, δ is strictly positive with boundary limit Δ.
    Lays out the working-region framework; all later theorems inherit this restricted setting, so they are local rather than global.
  • ad hoc to paper Theorem 9 uses 'contractibility of W' to conclude w2(TE_W)=0.
    A compact connected open subset of a manifold is not necessarily contractible; this assumption is unproved and is load-bearing for the existence of the symplectic form.
  • ad hoc to paper Assumption 2 and Lemma 27: Y can be treated as R^k via isometric Whitney embedding with Euler class preservation, assuming e(N)=0 for the normal bundle.
    Normal bundles of arbitrary submanifolds are not generally parallelizable; the reduction to Euclidean observation space is not established.
  • domain assumption Definition 6: safety potential Φ satisfies radial eventual monotonicity and stratified properness.
    These conditions are imposed rather than derived and are needed for the properness theorem and constraint surface regularity.
  • domain assumption Definition 7: observation-adaptive connection has vanishing vertical curvature, bounded mixed curvature, and degenerate boundary behavior.
    The closedness and non-degeneracy of the constructed symplectic form depend on these curvature assumptions.
  • domain assumption Remark 11: fiber metric is compatible with pullback metric h^*g_Y, ρ_{ab}=(h^*g_Y)_{ab}.
    Used to justify structure-function symmetries needed for dΩ_mix=0.
invented entities (3)
  • observation-induced fiber bundle E_W -> W
    purpose: Encodes all possible observation deviations ξ in cotangent spaces T*_{h(x)}Y as fibers over each state.
    Core construction of the paper; no independent observable signature is predicted beyond the framework itself.
  • observation-adaptive connection
    purpose: Defines horizontal-vertical splitting and determines mixed symplectic terms and Poisson brackets.
    Introduced to make the geometry work; assumptions about its curvature are not independently verifiable from data.
  • boundary degenerate symplectic form with safety buffer
    purpose: Provides controlled dissipation and inward attraction near the boundary of the working region.
    Hypothetical mechanism; no experimental measurement or simulation is provided in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Constrained Hamiltonian Systems on Observation-Induced Fiber Bundles: Theory of Symmetry and Integrability." pith.science (2026). https://pith.science/paper/JGYP45ZE

@misc{pith2026250522824,
  author       = {Pith},
  title        = {Pith review of: Constrained Hamiltonian Systems on Observation-Induced Fiber Bundles: Theory of Symmetry and Integrability},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JGYP45ZE}},
  note         = {Machine review of arXiv:2505.22824}
}
read the original abstract

Classical constrained Hamiltonian theory assumes complete observability of system states, but in reality only partial state information is often available. This paper establishes a complete geometric theoretical framework for handling such incompletely observed systems. By introducing the concept of observation-induced fiber bundles, we naturally extend Dirac constraint theory to the fiber bundle setting, unifying the treatment of state constraints and observation constraints. Main results include: (1) Classification of existence conditions for observation fiber bundles based on characteristic class theory; (2) Complete characterization of Poisson structures on fiber bundles and corresponding symplectic reduction theory; (3) Geometric necessary and sufficient conditions for integrability and Lax pair construction; (4) Extension of Noether's theorem under symmetry group actions. The theoretical framework naturally encompasses a wide range of applications from classical mechanics to modern safety-critical control systems, providing a rigorous mathematical foundation for dynamical analysis under incomplete information.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

26 extracted references · 25 canonical work pages

  1. [1]

    Information Geometry and Its Applications

    Shun-ichi Amari. Information Geometry and Its Applications. Springer, Tokyo, 2016

    work page 2016

  2. [2]

    Control barrier functions: Theory and applications

    Aaron D Ames, Samuel Coogan, Magnus Egerstedt, Gennaro Notomista, Koushil Sreenath, and Paulo Tabuada. Control barrier functions: Theory and applications. 18th European Control Conference (ECC), pages 3420–3431, 2019

    work page 2019

  3. [3]

    Mathematical methods of classical mechanics , volume 60

    Vladimir I Arnold. Mathematical methods of classical mechanics , volume 60. Springer Science & Business Media, 1989

    work page 1989

  4. [4]

    The Biomedical Engineering Handbook

    Joseph D Bronzino. The Biomedical Engineering Handbook. CRC Press, Boca Raton, FL, 2nd edition, 2000

    work page 2000

  5. [5]

    Lectures on quantum mechanics, volume 2

    Paul Adrien Maurice Dirac. Lectures on quantum mechanics, volume 2. Yeshiva University, 1964

    work page 1964

  6. [6]

    Generalized hamiltonian dynamics

    Paul AM Dirac. Generalized hamiltonian dynamics. Canadian journal of mathematics , 2(2):129–148, 1950

    work page 1950

  7. [7]

    Les connexions infinit ´esimales dans un espace fibr ´e diff ´erentiable

    Charles Ehresmann. Les connexions infinit ´esimales dans un espace fibr ´e diff ´erentiable. Colloque de topologie (espaces fibr´es), Bruxelles, pages 29–55, 1950

    work page 1950

  8. [8]

    On a general method in dynamics

    William R Hamilton. On a general method in dynamics. Philosophical Transactions of the Royal Society of London, 124:247–308, 1834

    work page

Show all 26 references
  1. [9]

    Quantization of Gauge Systems

    Marc Henneaux and Claudio Teitelboim. Quantization of Gauge Systems . Princeton University Press, Princeton, 1992

    1992

  2. [10]

    Vorlesungen ¨uber Dynamik

    Carl Gustav Jacob Jacobi. Vorlesungen ¨uber Dynamik. Reimer, 1866. 47

  3. [11]

    Stochastic Processes and Filtering Theory

    Andrew H Jazwinski. Stochastic Processes and Filtering Theory. Academic Press, New York, 1970

    1970

  4. [12]

    Nonlinear Systems

    Hassan K Khalil. Nonlinear Systems. Prentice Hall, Upper Saddle River, NJ, 3rd edition, 2002

    2002

  5. [13]

    Reduction of symplectic manifolds with symmetry

    Jerrold Marsden and Alan Weinstein. Reduction of symplectic manifolds with symmetry. Reports on mathematical physics, 5(1):121–130, 1974

    1974

  6. [14]

    A Mathematical Introduction to Robotic Manipulation

    Richard M Murray, Zexiang Li, and S Shankar Sastry. A Mathematical Introduction to Robotic Manipulation. CRC Press, Boca Raton, FL, 1994

    1994

  7. [15]

    Convex approximations of chance constrained programs

    Arkadi Nemirovski and Alexander Shapiro. Convex approximations of chance constrained programs. SIAM Journal on Optimization, 17(4):969–996, 2007

    2007

  8. [16]

    Cambridge University Press, Cambridge, 2000

    Michael A Nielsen and Isaac L Chuang.Quantum Computation and Quantum Information. Cambridge University Press, Cambridge, 2000

    2000

  9. [17]

    Numerical Optimization

    Jorge Nocedal and Stephen J Wright. Numerical Optimization. Springer, New York, 2nd edition, 2006

    2006

  10. [18]

    Invariante variationsprobleme

    Emmy Noether. Invariante variationsprobleme. Nachrichten von der Gesellschaft der Wissenschaften zu G¨ottingen, Mathematisch-Physikalische Klasse, 1918:235–257, 1918

    1918

  11. [19]

    Consensus and cooperation in networked multi-agent systems

    Reza Olfati-Saber, J Alex Fax, and Richard M Murray. Consensus and cooperation in networked multi-agent systems. Proceedings of the IEEE, 95(1):215–233, 2007

    2007

  12. [20]

    A survey of motion planning and control techniques for self-driving urban vehicles

    Brian Paden, Michal Cap, Sze Zheng Yong, Dmitry Yershov, and Emilio Frazzoli. A survey of motion planning and control techniques for self-driving urban vehicles. IEEE Transactions on Intelligent Vehicles, 1(1):33–55, 2016

    2016

  13. [21]

    Model Predictive Control: Theory, Computation, and Design

    James Blake Rawlings, David Q Mayne, and Moritz Diehl. Model Predictive Control: Theory, Computation, and Design. Nob Hill Publishing, Madison, WI, 2nd edition, 2017

    2017

  14. [22]

    The Topology of Fibre Bundles

    Norman Earl Steenrod. The Topology of Fibre Bundles . Princeton University Press, Princeton, 1951

    1951

  15. [23]

    Aircraft Control and Simulation: Dynamics, Controls Design, and Autonomous Systems

    Brian L Stevens, Frank L Lewis, and Eric N Johnson. Aircraft Control and Simulation: Dynamics, Controls Design, and Autonomous Systems. John Wiley & Sons, Hoboken, NJ, 3rd edition, 2015

    2015

  16. [24]

    Conservation of isotopic spin and isotopic gauge invariance

    Chen-Ning Yang and Robert L Mills. Conservation of isotopic spin and isotopic gauge invariance. Physical Review, 96(1):191–195, 1954

    1954

  17. [25]

    Learning dynamics under environmental constraints via measurement-induced bundle structures

    Dongzhe Zheng and Wenjie Mei. Learning dynamics under environmental constraints via measurement-induced bundle structures. In Proceedings of the 41st International Conference on Machine Learning. PMLR, 2025

    2025

  18. [26]

    Prentice Hall, Upper Saddle River, NJ, 1996

    Kemin Zhou, John C Doyle, and Keith Glover.Robust and Optimal Control. Prentice Hall, Upper Saddle River, NJ, 1996. 48

    1996

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.