Risk-sensitive linear-quadratic-Gaussian graphon mean-field games

Minyi Huang; Tian Chen

arxiv: 2604.23313 · v1 · submitted 2026-04-25 · 🧮 math.OC

Risk-sensitive linear-quadratic-Gaussian graphon mean-field games

Tian Chen , Minyi Huang This is my paper

Pith reviewed 2026-05-08 07:34 UTC · model grok-4.3

classification 🧮 math.OC

keywords risk-sensitive controlgraphon mean-field gameslinear-quadratic-Gaussianepsilon-Nash equilibriumdecentralized strategiesforward-backward differential equations

0 comments

The pith

Decentralized strategies derived from graphon mean-field equations satisfy the epsilon-Nash property for risk-sensitive LQG agents.

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

This paper constructs decentralized strategies for an asymptotically infinite population of heterogeneous agents on an asymptotically infinite network, where each agent minimizes an exponential cost that encodes risk sensitivity. It does so by solving an auxiliary risk-sensitive optimal control problem together with a consistency condition, which yields a family of fully coupled forward-backward differential equations. The system is shown to be well-posed either by a contraction mapping or by the method of continuity. The resulting strategies are then proved to form an epsilon-Nash equilibrium of the original game through exponentiated error bounds that accommodate the unbounded integrand in the exponential cost. A sympathetic reader would care because the result supplies a tractable, decentralized approximation for risk-averse decision making in large networked systems without requiring a central coordinator.

Core claim

The graphon mean-field game equation system consisting of fully coupled forward-backward differential equations is well-posed under a contraction condition or an operator monotonicity condition. The decentralized strategies obtained from this system possess the epsilon-Nash equilibrium property for the infinite-population game, established by deriving exponentiated error estimates rather than the usual L2 estimates.

What carries the argument

The graphon mean-field game equation system, a family of fully coupled forward-backward differential equations obtained via Nash certainty equivalence and a consistency condition, which produces the decentralized strategies.

If this is right

The obtained strategies remain approximately optimal when the population and network are replaced by large but finite versions.
Exponentiated error estimates replace L2 estimates whenever the cost integrand is unbounded.
Well-posedness of the equation system guarantees existence of the decentralized controls under the stated conditions.
A numerical example confirms that the strategies can be computed and perform as predicted.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same exponentiated-error technique could be tested on graphon games with non-quadratic running costs.
For finite but large networks the graphon limit supplies an explicit rate at which the epsilon-Nash gap shrinks.
Risk-sensitive graphon games may serve as a tractable test bed for studying how network topology interacts with exponential utility.

Load-bearing premise

A contraction condition or an operator monotonicity condition holds for the family of fully coupled forward-backward differential equations.

What would settle it

A specific choice of risk-sensitivity parameter and graphon for which the fixed-point map fails to be contractive, yet numerical simulation of finite-agent trajectories shows that the realized cost deviation remains bounded by a vanishing epsilon.

Figures

Figures reproduced from arXiv: 2604.23313 by Minyi Huang, Tian Chen.

**Figure 1.** Figure 1: Graphon and its section. 0 0.2 0.4 0.6 0.8 1 time 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 value view at source ↗

**Figure 2.** Figure 2: The solutions (Π, P ⊥, P 1 2 , P 1 4 ) to Riccati equations. (a) State ¯xα (b) Control ¯uα view at source ↗

**Figure 3.** Figure 3: The trajectories of state and control for 200 different α values under the graphon limit. For illustration, we now consider a sinusoidal graphon as follows (6.4) G(α, β) = cos2 π 2 (α − β) , α, β ∈ [0, 1], view at source ↗

**Figure 4.** Figure 4: The trajectories of the state and control with 4 different α values. 0 0.2 0.4 0.6 0.8 1 time 1 1.5 2 2.5 3 3.5 4 4.5 5 value (a) Graphon weighted mean state zα 0 0.2 0.4 0.6 0.8 1 time 2 3 4 5 6 7 8 9 10 11 value (b) ODE solution Sα view at source ↗

**Figure 5.** Figure 5: The solutions to the GMFG equation system (4.6). which is illustrated in Fig. 1a. For this sinusoidal graphon G, we can show that the normalized eigenfunctions are f1 = 1, f2 = √ 2 cos π(·) and f3 = √ 2 sin π(·) with eigenvalues λ1 = 1 2 , λ2 = λ3 = 1 4 . We take parameters T = 1, γ = 0.3, A = 0.5, B = 0.6, D = 2, σ = 0.5, Q = 0.3, Γ = 2, R = 1.5, Qf = 0.8, Γf = −0.8. The numerical solutions of (Π, P ⊥, P … view at source ↗

read the original abstract

This paper investigates a class of linear-quadratic-Gaussian risk-sensitive graphon mean-field games, involving an asymptotically infinite population of heterogeneous agents distributed across an asymptotically infinite network, where each agent aims to minimize an exponential cost functional reflecting its risk sensitivity. Following the Nash certainty equivalence methodology, an auxiliary risk-sensitive optimal control problem is constructed and further combined with a consistency condition to determine decentralized strategies of the agents. The well-posedness of the resulting graphon mean-field game equation system, consisting of a family of fully coupled forward-backward differential equations, is established by a fixed point approach under a contraction condition, and by the method of continuity under an operator monotonicity condition, respectively. To prove the epsilon-Nash equilibrium property of the obtained decentralized strategies, one faces significant challenge since the usual L^2 error estimates on mean-field approximations are no longer adequate due to unboundedness of the integrand in the exponentiated cost. The proof will be accomplished by establishing certain exponentiated error estimates instead of L^2 error estimates. Finally, a numerical example is provided to illustrate our results.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a workable extension of risk-sensitive LQG to graphon mean-field games via exponentiated error estimates, but the well-posedness of the core FBDE system rests on unverified contraction or monotonicity conditions.

read the letter

The main thing to know is that they adapt the usual Nash certainty equivalence setup to risk-sensitive costs on graphon networks and replace standard L2 error bounds with exponentiated ones to handle the fact that the integrand in the cost is unbounded. That switch lets them claim an epsilon-Nash property for the decentralized strategies once the mean-field system is solved. The combination itself and the error technique look new relative to the plain LQG graphon MFG papers that came before it. They also include a numerical example, which is helpful for seeing the construction in action at least in one case. The abstract lays out a coherent plan: build the auxiliary control problem, impose the consistency condition, prove well-posedness of the resulting family of fully coupled forward-backward equations either by contraction mapping or by the method of continuity, then run the exponentiated estimates to get the equilibrium property. That structure is straightforward and stays within the existing literature's framework without obvious circularity. The soft spot is exactly where the stress-test note points. Both routes to well-posedness require either a contraction condition or an operator monotonicity condition, yet the abstract gives no parameter regime, no smallness assumption on the risk-sensitivity parameter, and no graphon-dependent bound that would guarantee these hold. If the conditions fail for some graphons or risk levels, the candidate strategies are not even defined and the later error argument cannot be applied. The paper states the conditions explicitly but does not close this gap with explicit verification or sufficient conditions, so the central claim is conditional on something that remains unchecked in general. This work is for people already working in mean-field games, stochastic control, and network games who need to incorporate risk aversion. A reader in that subfield would pick up the exponentiated estimate idea and the graphon extension even if they have to supply their own smallness assumptions later. It is technically plausible and addresses a real obstacle with unbounded costs, so it deserves a serious referee rather than a desk reject. I would send it out for review with the expectation that the authors will need to add concrete conditions or counter-examples showing when the fixed-point or continuity arguments actually apply.

Referee Report

1 major / 1 minor

Summary. This paper investigates risk-sensitive linear-quadratic-Gaussian graphon mean-field games for an asymptotically infinite population of heterogeneous agents on an asymptotically infinite network. Each agent minimizes an exponential cost functional. Following Nash certainty equivalence, an auxiliary risk-sensitive optimal control problem is combined with a consistency condition to derive decentralized strategies. Well-posedness of the resulting family of fully coupled forward-backward differential equations is established via a fixed-point approach under a contraction condition and via the method of continuity under an operator monotonicity condition. The epsilon-Nash equilibrium property of these strategies is proved using exponentiated error estimates (rather than standard L2 estimates) to handle the unbounded integrand in the cost functional. A numerical example is provided to illustrate the results.

Significance. If the contraction or monotonicity conditions hold under verifiable assumptions, the work makes a meaningful contribution to graphon mean-field game theory by extending it to the risk-sensitive LQG setting with heterogeneous agents and networks. The shift to exponentiated error estimates is a technically appropriate response to the failure of L2 bounds for exponential costs and represents a strength of the approach. The numerical example adds concrete support, and the overall framework could inform applications in large-scale networked control with risk aversion.

major comments (1)

[Abstract / well-posedness section] Abstract and well-posedness analysis: well-posedness of the graphon mean-field game equation system (the family of fully coupled FBDEs) is established only under a contraction condition (fixed-point approach) or operator monotonicity condition (method of continuity). No parameter regime, smallness assumption on the risk-sensitivity parameter, or graphon-dependent bound is supplied to guarantee these conditions hold for the LQG risk-sensitive case. This is load-bearing for the central claim, because without solutions to the system the decentralized strategies are undefined and the subsequent exponentiated-error argument for the epsilon-Nash property cannot be invoked.

minor comments (1)

[Numerical example] The numerical example would benefit from explicit statements of the chosen graphon, the values of the risk-sensitivity parameter, and the discretization scheme to facilitate reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and constructive feedback on our manuscript. We address the major comment below.

read point-by-point responses

Referee: [Abstract / well-posedness section] Abstract and well-posedness analysis: well-posedness of the graphon mean-field game equation system (the family of fully coupled FBDEs) is established only under a contraction condition (fixed-point approach) or operator monotonicity condition (method of continuity). No parameter regime, smallness assumption on the risk-sensitivity parameter, or graphon-dependent bound is supplied to guarantee these conditions hold for the LQG risk-sensitive case. This is load-bearing for the central claim, because without solutions to the system the decentralized strategies are undefined and the subsequent exponentiated-error argument for the epsilon-Nash property cannot be invoked.

Authors: We agree that the well-posedness results are conditional on the contraction or operator monotonicity assumptions, and that the manuscript does not supply explicit parameter regimes, smallness conditions on the risk-sensitivity parameter, or graphon-dependent bounds guaranteeing these assumptions in the LQG risk-sensitive setting. Deriving such general, verifiable bounds for arbitrary graphons is technically involved and lies outside the scope of the present work; the paper instead focuses on establishing the equilibrium equations and the epsilon-Nash property whenever the well-posedness conditions hold. This conditional approach is standard in the mean-field game literature. The numerical example demonstrates practical convergence of the fixed-point iteration for a concrete graphon and risk parameter, indicating that the conditions can be satisfied in applications. In the revised version we will add a short discussion in the well-posedness section clarifying the conditional nature of the results and noting that the conditions may be verified numerically or for specific graphons, thereby addressing the concern that the decentralized strategies require existence of solutions to the FBDE system. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper follows the standard Nash certainty equivalence approach to construct an auxiliary optimal control problem plus consistency condition, then establishes well-posedness of the resulting family of fully coupled forward-backward equations under explicit contraction or operator-monotonicity assumptions, and finally proves the epsilon-Nash property via exponentiated error estimates. None of these steps reduce by construction to the paper's own inputs, fitted parameters renamed as predictions, or self-referential definitions. The methodology is drawn from prior literature (not shown to be load-bearing self-citation within this manuscript), and the central consistency condition plus error estimates retain independent mathematical content. No quoted equation or derivation collapses to a tautology or unverified self-citation chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work rests on standard domain assumptions from mean-field game theory and stochastic control; without the full manuscript, specific free parameters or invented entities cannot be enumerated beyond the graphon measurability and boundedness implicitly required for the consistency condition.

axioms (1)

domain assumption The graphon is measurable and bounded
Required for the graphon mean-field limit and consistency condition to be well-defined

pith-pipeline@v0.9.0 · 5483 in / 1181 out tokens · 37638 ms · 2026-05-08T07:34:52.399395+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

47 extracted references · 47 canonical work pages

[1]

Abou-Kandil, G

H. Abou-Kandil, G. Freiling, V. Ionescu, and G. Jank, Matrix riccati equations in control and systems theory , Springer Basel AG, Basel, Switzerland, 2003

work page 2003
[2]

Aurell, R

A. Aurell, R. Carmona, and M. Lauriere, Stochastic graphon games: II. the linear-quadratic case, Appl. Math. Optim. 85 (2022), no. 3, 39

work page 2022
[3]

Ba¸ sar and G

T. Ba¸ sar and G. J. Olsder, Dynamic noncooperative game theory, vol. 23, SIAM, Philadelphia, PA, 1998

work page 1998
[4]

Bauso, X

D. Bauso, X. Zhang, and A. Papachristodoulou, Density ﬂow in dynamical networks via mean-ﬁeld games , IEEE Trans. Autom. Control 62 (2017), no. 3, 1342–1355

work page 2017
[5]

Bayraktar, S

E. Bayraktar, S. Chakraborty, and R. W u, Graphon mean ﬁeld systems , Ann. Appl. Probab. 33 (2023), no. 5, 3587–3619

work page 2023
[6]

Bensoussan, J

A. Bensoussan, J. Frehse, and H. Nagai, Some results on risk-sensitive control with full ob- servation, Appl. Math. Optim. 37 (1998), 1–41

work page 1998
[7]

Bensoussan, J

A. Bensoussan, J. Frehse, and P. Yam, Mean ﬁeld games and mean ﬁeld type control theory , vol. 101, Springer, New York, USA, 2013

work page 2013
[8]

Borgs, J

C. Borgs, J. T. Chayes, L. Lov´ asz, V. T. S´ os, and K. Veszte rgombi, Convergent sequences of dense graphs I: Subgraph frequencies, metric properties an d testing , Adv. Math. 219 (2008), no. 6, 1801–1851

work page 2008
[9]

Multiway cuts and s tatistical physics, Ann

, Convergent sequences of dense graphs II. Multiway cuts and s tatistical physics, Ann. Math. (2012), 151–219

work page 2012
[10]

P. E. Caines and M. Huang, Graphon mean ﬁeld games and their equations , SIAM J. Control Optim. 59 (2021), no. 6, 4373–4399

work page 2021
[11]

Cardaliaguet, F

P. Cardaliaguet, F. Delarue, J. M. Lasry, and P. L. Lions, The master equation and the convergence problem in mean ﬁeld games:(AMS-201) , Princeton University Press, 2019

work page 2019
[12]

Carmona, D

R. Carmona, D. B. Cooney, C. V. Graves, and M. Lauriere, Stochastic graphon games: I. the static case , Math. Oper. Res. 47 (2022), no. 1, 750–778

work page 2022
[13]

Carmona and F

R. Carmona and F. Delarue, Probabilistic theory of mean ﬁeld games with applications I -II, Springer, New York, USA, 2018

work page 2018
[14]

Y. Chen, T. Li, and Z. Xin, Risk-sensitive mean ﬁeld games with major and minor players , ESAIM-Control Optim. Calc. Var. 29 (2023), 6

work page 2023
[15]

S. N. Cohen and R. J. Elliott, Stochastic calculus and applications , vol. 2, Springer, New York, 2015

work page 2015
[16]

Delarue, Mean ﬁeld games: A toy model on an Erd¨ os-Renyi graph

F. Delarue, Mean ﬁeld games: A toy model on an Erd¨ os-Renyi graph. , ESAIM. Proc. Surv. 60 (2017), 1–26

work page 2017
[17]

T. E. Duncan, Linear-exponential-quadratic Gaussian control , IEEE Trans. Autom. Control 58 (2013), no. 11, 2910–2911

work page 2013
[18]

G. E. Espinosa and N. Touzi, Optimal investment under relative performance concerns , Math. Financ. 25 (2015), no. 2, 221–257

work page 2015
[19]

W. H. Fleming and D. Hern´ andez-Hern´ andez, Risk-sensitive control of ﬁnite state machines on an inﬁnite horizon I , SIAM J. Control Optim. 35 (1997), no. 5, 1790–1810. 32 TIAN CHEN AND MINYI HUANG

work page 1997
[20]

W. H. Fleming and W. M. McEneaney, Risk-sensitive control on an inﬁnite time horizon , SIAM J. Control Optim. 33 (1995), no. 6, 1881–1915

work page 1995
[21]

Foguen-Tchuendom, S

R. Foguen-Tchuendom, S. Gao, P. E. Caines, and M. Huang, Inﬁnite horizon LQG graphon mean ﬁeld games: Explicit Nash values and local minima , Syst. Control Lett. 187 (2024), 105780

work page 2024
[22]

Fu and C

G. Fu and C. Zhou, Mean ﬁeld portfolio games , Financ. Stoch. 27 (2023), no. 1, 189–231

work page 2023
[23]

S. Gao, P. E. Caines, and M. Huang, LQG graphon mean ﬁeld games: Analysis via graphon- invariant subspaces, IEEE Trans. Autom. Control 68 (2023), no. 12, 7482–7497

work page 2023
[24]

Howard and J.E

R.A. Howard and J.E. Matheson, Risk-sensitive Markov decision processes , Manag. Sci. 18 (1972), no. 7, 356–369

work page 1972
[25]

Hu and S

Y. Hu and S. Peng, Solution of forward-backward stochastic diﬀerential equa tions, Probab. Theory Related Fields 103 (1995), 273–283

work page 1995
[26]

Huang, P

M. Huang, P. E. Caines, and R. P. Malham´ e, Large-population cost-coupled LQG problems with nonuniform agents: Individual-mass behavior and dece ntralized ε-Nash equilibria , IEEE Trans. Autom. Control 52 (2007), no. 9, 1560–1571

work page 2007
[27]

Huang, R.P

M. Huang, R.P. Malham´ e, and P.E. Caines, Large population stochastic dynamic games: Closed-loop McKean-Vlasov systems and the Nash certainty e quivalence principle , Commun. Inf. Syst. 6 (2006), no. 1, 221–252

work page 2006
[28]

Huang and M

M. Huang and M. Zhou, Linear quadratic mean ﬁeld games: Asymptotic solvability a nd relation to the ﬁxed point approach , IEEE Trans. Autom. Control 65 (2020), no. 4, 1397– 1412

work page 2020
[29]

Jacobson, Optimal stochastic linear systems with exponential perfor mance criteria and their relation to deterministic diﬀerential games , IEEE Trans

D. Jacobson, Optimal stochastic linear systems with exponential perfor mance criteria and their relation to deterministic diﬀerential games , IEEE Trans. Autom. Control 18 (1973), no. 2, 124–131

work page 1973
[30]

Lacker, A general characterization of the mean ﬁeld limit for stocha stic diﬀerential games , Probab

D. Lacker, A general characterization of the mean ﬁeld limit for stocha stic diﬀerential games , Probab. Theory Relat. Field 165 (2016), 581–648

work page 2016
[31]

J. M. Lasry and P. L. Lions, Mean ﬁeld games , Jpn. J. Math. 2 (2007), no. 1, 229–260

work page 2007
[32]

A. E. B. Lim and X. Y. Zhou, A new risk-sensitive maximum principle , IEEE Trans. Autom. Control 50 (2005), no. 7, 958–966

work page 2005
[33]

H. Liu, D. Firoozi, and M. Breton, LQG risk-sensitive single-agent and major-minor mean ﬁeld game systems: A variational framework , SIAM J. Control Optim., in press 63 (2025), no. 4, 2251–2281

work page 2025
[34]

Lov´ asz, Large networks and graph limits , vol

L. Lov´ asz, Large networks and graph limits , vol. 60, American Mathematical Society, 2012

work page 2012
[35]

Lov´ asz and B

L. Lov´ asz and B. Szegedy,Limits of dense graph sequences , J. Comb. Theory Ser. B 96 (2006), no. 6, 933–957

work page 2006
[36]

Mathai and S.B

A.M. Mathai and S.B. Provost, Quadratic forms in random variables: Theory and applica- tions, Marcel Dekker, INC, New York, USA, 1992

work page 1992
[37]

Moon and T

J. Moon and T. Ba¸ sar, Risk-sensitive mean ﬁeld games via the stochastic maximum p rinciple, Dyn. Games Appl. 9 (2019), no. 4, 1100–1125

work page 2019
[38]

Pan and T

Z. Pan and T. Ba¸ sar, Model simpliﬁcation and optimal control of stochastic sing ularly per- turbed systems under exponentiated quadratic cost , SIAM J. Control Optim. 34 (1996), no. 5, 1734–1766

work page 1996
[39]

Pazy, Semigroups of linear operators and applications to partial diﬀerential equations , vol

A. Pazy, Semigroups of linear operators and applications to partial diﬀerential equations , vol. 44, Springer Science & Business Media, New York, USA, 20 12

work page
[40]

Peng and Z

S. Peng and Z. W u, Fully coupled forward-backward stochastic diﬀerential eq uations and ap- plications to optimal control , SIAM J. Control Optim. 37 (1999), no. 3, 825–843

work page 1999
[41]

Saldi, T

N. Saldi, T. Ba¸ sar, and M. Raginsky, Approximate Markov-Nash equilibria for discrete-time risk-sensitive mean-ﬁeld games , Math. Oper. Res. 45 (2020), no. 4, 1596–1620

work page 2020
[42]

R. E. Showalter, Monotone operators in banach space and nonlinear partial di ﬀerential equa- tions, vol. 49, American Mathematical Society, 2013

work page 2013
[43]

Tembine, Q Zhu, and T

H. Tembine, Q Zhu, and T. Ba¸ sar, Risk-sensitive mean-ﬁeld games , IEEE Trans. Autom. Control 59 (2014), no. 4, 835–850

work page 2014
[44]

W ang and M

B. W ang and M. Huang, Mean ﬁeld production output control with sticky prices: Nas h and social solutions , Automatica 100 (2019), 90–98

work page 2019
[45]

W ang and M

Y. W ang and M. Huang, Risk-sensitive linear-quadratic mean-ﬁeld games: Asympt otic solv- ability and decentralized O(1/N )-Nash equilibria , J. Syst. Sci. Complex. 38 (2025), no. 1, 436–459. RISK-SENSITIVE LINEAR-QUADRATIC-GAUSSIAN GRAPHON MEAN- FIELD GAMES 33

work page 2025
[46]

Whittle, Risk-sensitive optimal control , American Mathematical Society, New York, USA, 1990

P. Whittle, Risk-sensitive optimal control , American Mathematical Society, New York, USA, 1990

work page 1990
[47]

Yosida, Functional analysis , 6 ed., Springer Science, Berlin Heidelberg, 1980

K. Yosida, Functional analysis , 6 ed., Springer Science, Berlin Heidelberg, 1980. School of Mathematics, Shandong University, Jinan, 250100, China; and School of Mathematics and Statistics, Carleton University, Ottaw a, O N K1S 5B6, Canada. Email address : chentian43@sdu.edu.cn School of Mathematics and Statistics, Carleton University, Ottaw a, ON K1S 5B...

work page 1980

[1] [1]

Abou-Kandil, G

H. Abou-Kandil, G. Freiling, V. Ionescu, and G. Jank, Matrix riccati equations in control and systems theory , Springer Basel AG, Basel, Switzerland, 2003

work page 2003

[2] [2]

Aurell, R

A. Aurell, R. Carmona, and M. Lauriere, Stochastic graphon games: II. the linear-quadratic case, Appl. Math. Optim. 85 (2022), no. 3, 39

work page 2022

[3] [3]

Ba¸ sar and G

T. Ba¸ sar and G. J. Olsder, Dynamic noncooperative game theory, vol. 23, SIAM, Philadelphia, PA, 1998

work page 1998

[4] [4]

Bauso, X

D. Bauso, X. Zhang, and A. Papachristodoulou, Density ﬂow in dynamical networks via mean-ﬁeld games , IEEE Trans. Autom. Control 62 (2017), no. 3, 1342–1355

work page 2017

[5] [5]

Bayraktar, S

E. Bayraktar, S. Chakraborty, and R. W u, Graphon mean ﬁeld systems , Ann. Appl. Probab. 33 (2023), no. 5, 3587–3619

work page 2023

[6] [6]

Bensoussan, J

A. Bensoussan, J. Frehse, and H. Nagai, Some results on risk-sensitive control with full ob- servation, Appl. Math. Optim. 37 (1998), 1–41

work page 1998

[7] [7]

Bensoussan, J

A. Bensoussan, J. Frehse, and P. Yam, Mean ﬁeld games and mean ﬁeld type control theory , vol. 101, Springer, New York, USA, 2013

work page 2013

[8] [8]

Borgs, J

C. Borgs, J. T. Chayes, L. Lov´ asz, V. T. S´ os, and K. Veszte rgombi, Convergent sequences of dense graphs I: Subgraph frequencies, metric properties an d testing , Adv. Math. 219 (2008), no. 6, 1801–1851

work page 2008

[9] [9]

Multiway cuts and s tatistical physics, Ann

, Convergent sequences of dense graphs II. Multiway cuts and s tatistical physics, Ann. Math. (2012), 151–219

work page 2012

[10] [10]

P. E. Caines and M. Huang, Graphon mean ﬁeld games and their equations , SIAM J. Control Optim. 59 (2021), no. 6, 4373–4399

work page 2021

[11] [11]

Cardaliaguet, F

P. Cardaliaguet, F. Delarue, J. M. Lasry, and P. L. Lions, The master equation and the convergence problem in mean ﬁeld games:(AMS-201) , Princeton University Press, 2019

work page 2019

[12] [12]

Carmona, D

R. Carmona, D. B. Cooney, C. V. Graves, and M. Lauriere, Stochastic graphon games: I. the static case , Math. Oper. Res. 47 (2022), no. 1, 750–778

work page 2022

[13] [13]

Carmona and F

R. Carmona and F. Delarue, Probabilistic theory of mean ﬁeld games with applications I -II, Springer, New York, USA, 2018

work page 2018

[14] [14]

Y. Chen, T. Li, and Z. Xin, Risk-sensitive mean ﬁeld games with major and minor players , ESAIM-Control Optim. Calc. Var. 29 (2023), 6

work page 2023

[15] [15]

S. N. Cohen and R. J. Elliott, Stochastic calculus and applications , vol. 2, Springer, New York, 2015

work page 2015

[16] [16]

Delarue, Mean ﬁeld games: A toy model on an Erd¨ os-Renyi graph

F. Delarue, Mean ﬁeld games: A toy model on an Erd¨ os-Renyi graph. , ESAIM. Proc. Surv. 60 (2017), 1–26

work page 2017

[17] [17]

T. E. Duncan, Linear-exponential-quadratic Gaussian control , IEEE Trans. Autom. Control 58 (2013), no. 11, 2910–2911

work page 2013

[18] [18]

G. E. Espinosa and N. Touzi, Optimal investment under relative performance concerns , Math. Financ. 25 (2015), no. 2, 221–257

work page 2015

[19] [19]

W. H. Fleming and D. Hern´ andez-Hern´ andez, Risk-sensitive control of ﬁnite state machines on an inﬁnite horizon I , SIAM J. Control Optim. 35 (1997), no. 5, 1790–1810. 32 TIAN CHEN AND MINYI HUANG

work page 1997

[20] [20]

W. H. Fleming and W. M. McEneaney, Risk-sensitive control on an inﬁnite time horizon , SIAM J. Control Optim. 33 (1995), no. 6, 1881–1915

work page 1995

[21] [21]

Foguen-Tchuendom, S

R. Foguen-Tchuendom, S. Gao, P. E. Caines, and M. Huang, Inﬁnite horizon LQG graphon mean ﬁeld games: Explicit Nash values and local minima , Syst. Control Lett. 187 (2024), 105780

work page 2024

[22] [22]

Fu and C

G. Fu and C. Zhou, Mean ﬁeld portfolio games , Financ. Stoch. 27 (2023), no. 1, 189–231

work page 2023

[23] [23]

S. Gao, P. E. Caines, and M. Huang, LQG graphon mean ﬁeld games: Analysis via graphon- invariant subspaces, IEEE Trans. Autom. Control 68 (2023), no. 12, 7482–7497

work page 2023

[24] [24]

Howard and J.E

R.A. Howard and J.E. Matheson, Risk-sensitive Markov decision processes , Manag. Sci. 18 (1972), no. 7, 356–369

work page 1972

[25] [25]

Hu and S

Y. Hu and S. Peng, Solution of forward-backward stochastic diﬀerential equa tions, Probab. Theory Related Fields 103 (1995), 273–283

work page 1995

[26] [26]

Huang, P

M. Huang, P. E. Caines, and R. P. Malham´ e, Large-population cost-coupled LQG problems with nonuniform agents: Individual-mass behavior and dece ntralized ε-Nash equilibria , IEEE Trans. Autom. Control 52 (2007), no. 9, 1560–1571

work page 2007

[27] [27]

Huang, R.P

M. Huang, R.P. Malham´ e, and P.E. Caines, Large population stochastic dynamic games: Closed-loop McKean-Vlasov systems and the Nash certainty e quivalence principle , Commun. Inf. Syst. 6 (2006), no. 1, 221–252

work page 2006

[28] [28]

Huang and M

M. Huang and M. Zhou, Linear quadratic mean ﬁeld games: Asymptotic solvability a nd relation to the ﬁxed point approach , IEEE Trans. Autom. Control 65 (2020), no. 4, 1397– 1412

work page 2020

[29] [29]

Jacobson, Optimal stochastic linear systems with exponential perfor mance criteria and their relation to deterministic diﬀerential games , IEEE Trans

D. Jacobson, Optimal stochastic linear systems with exponential perfor mance criteria and their relation to deterministic diﬀerential games , IEEE Trans. Autom. Control 18 (1973), no. 2, 124–131

work page 1973

[30] [30]

Lacker, A general characterization of the mean ﬁeld limit for stocha stic diﬀerential games , Probab

D. Lacker, A general characterization of the mean ﬁeld limit for stocha stic diﬀerential games , Probab. Theory Relat. Field 165 (2016), 581–648

work page 2016

[31] [31]

J. M. Lasry and P. L. Lions, Mean ﬁeld games , Jpn. J. Math. 2 (2007), no. 1, 229–260

work page 2007

[32] [32]

A. E. B. Lim and X. Y. Zhou, A new risk-sensitive maximum principle , IEEE Trans. Autom. Control 50 (2005), no. 7, 958–966

work page 2005

[33] [33]

H. Liu, D. Firoozi, and M. Breton, LQG risk-sensitive single-agent and major-minor mean ﬁeld game systems: A variational framework , SIAM J. Control Optim., in press 63 (2025), no. 4, 2251–2281

work page 2025

[34] [34]

Lov´ asz, Large networks and graph limits , vol

L. Lov´ asz, Large networks and graph limits , vol. 60, American Mathematical Society, 2012

work page 2012

[35] [35]

Lov´ asz and B

L. Lov´ asz and B. Szegedy,Limits of dense graph sequences , J. Comb. Theory Ser. B 96 (2006), no. 6, 933–957

work page 2006

[36] [36]

Mathai and S.B

A.M. Mathai and S.B. Provost, Quadratic forms in random variables: Theory and applica- tions, Marcel Dekker, INC, New York, USA, 1992

work page 1992

[37] [37]

Moon and T

J. Moon and T. Ba¸ sar, Risk-sensitive mean ﬁeld games via the stochastic maximum p rinciple, Dyn. Games Appl. 9 (2019), no. 4, 1100–1125

work page 2019

[38] [38]

Pan and T

Z. Pan and T. Ba¸ sar, Model simpliﬁcation and optimal control of stochastic sing ularly per- turbed systems under exponentiated quadratic cost , SIAM J. Control Optim. 34 (1996), no. 5, 1734–1766

work page 1996

[39] [39]

Pazy, Semigroups of linear operators and applications to partial diﬀerential equations , vol

A. Pazy, Semigroups of linear operators and applications to partial diﬀerential equations , vol. 44, Springer Science & Business Media, New York, USA, 20 12

work page

[40] [40]

Peng and Z

S. Peng and Z. W u, Fully coupled forward-backward stochastic diﬀerential eq uations and ap- plications to optimal control , SIAM J. Control Optim. 37 (1999), no. 3, 825–843

work page 1999

[41] [41]

Saldi, T

N. Saldi, T. Ba¸ sar, and M. Raginsky, Approximate Markov-Nash equilibria for discrete-time risk-sensitive mean-ﬁeld games , Math. Oper. Res. 45 (2020), no. 4, 1596–1620

work page 2020

[42] [42]

R. E. Showalter, Monotone operators in banach space and nonlinear partial di ﬀerential equa- tions, vol. 49, American Mathematical Society, 2013

work page 2013

[43] [43]

Tembine, Q Zhu, and T

H. Tembine, Q Zhu, and T. Ba¸ sar, Risk-sensitive mean-ﬁeld games , IEEE Trans. Autom. Control 59 (2014), no. 4, 835–850

work page 2014

[44] [44]

W ang and M

B. W ang and M. Huang, Mean ﬁeld production output control with sticky prices: Nas h and social solutions , Automatica 100 (2019), 90–98

work page 2019

[45] [45]

W ang and M

Y. W ang and M. Huang, Risk-sensitive linear-quadratic mean-ﬁeld games: Asympt otic solv- ability and decentralized O(1/N )-Nash equilibria , J. Syst. Sci. Complex. 38 (2025), no. 1, 436–459. RISK-SENSITIVE LINEAR-QUADRATIC-GAUSSIAN GRAPHON MEAN- FIELD GAMES 33

work page 2025

[46] [46]

Whittle, Risk-sensitive optimal control , American Mathematical Society, New York, USA, 1990

P. Whittle, Risk-sensitive optimal control , American Mathematical Society, New York, USA, 1990

work page 1990

[47] [47]

Yosida, Functional analysis , 6 ed., Springer Science, Berlin Heidelberg, 1980

K. Yosida, Functional analysis , 6 ed., Springer Science, Berlin Heidelberg, 1980. School of Mathematics, Shandong University, Jinan, 250100, China; and School of Mathematics and Statistics, Carleton University, Ottaw a, O N K1S 5B6, Canada. Email address : chentian43@sdu.edu.cn School of Mathematics and Statistics, Carleton University, Ottaw a, ON K1S 5B...

work page 1980