Pith. sign in

REVIEW 4 major objections 5 minor 59 references

The paper claims that portfolio skewness and kurtosis can be read as counts of balanced triangles and a specific signed 4-clique in daily signed graphs, and that hedge-score screening selects reduced asset universes that outperform the full

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A hedge-score signed-network method selects smaller stock sets that backtests show outperform the full universe, but the claimed higher-moment optimization is not implemented or validated.

T0 review reviewed 2026-08-02 challenge →

load-bearing objection The paper's central claim linking skewness to balanced triangles is false: in this graph every triangle is balanced, and the triple-product sign is not controlled by balance; the NP-hard proof also has a gap, and the backtest lacks proper baselines. the 4 major comments →

arxiv 2602.21362 v2 pith:CTHCC4BM submitted 2026-02-24 math.CO cs.CE

Signed network models for dimensionality reduction of portfolio optimization

Bibhas Adhikari This is my paper
classification math.CO cs.CE MSC 05C2291G1090C27
keywords signed graphportfolio optimizationdimensionality reductionhigher-order momentsskewnesskurtosishedge scorestructural balance
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 show that the higher-order moments of portfolio returns—skewness and kurtosis—have a direct combinatorial counterpart in a daily signed graph built from asset return co-movement relative to each asset's own mean. It argues that maximizing skewness corresponds to maximizing balanced triangles, and minimizing kurtosis corresponds to maximizing a particular balanced 4-clique pattern, and that the latter is NP-hard. On the practical side, the paper introduces a hedge score for each stock—the fraction of days it moves opposite to each other asset—and proposes selecting the top K stocks by hedge score times mean return. Backtesting on 199 S&P 500 assets from 2006 to 2021 shows that reduced universes of 20–50 stocks deliver moderately higher annual returns than the full universe in almost all test years, which matters because it offers a parameter-free, combinatorial route to dimension reduction that bypasses noisy correlation matrices.

Core claim

On its own terms, the paper's central discovery is a dictionary between portfolio moments and signed-graph patterns. For the skewness expression, the term involving three distinct assets is positive exactly when the triangle of their deviations in the daily signed graph is balanced (type T0 or T2); the paper proves the daily signed graph is always balanced, so skewness is naturally maximized by any complete subgraph. For kurtosis, the term involving four distinct assets is negative—hence kurtosis-reducing—exactly when the induced 4-clique is isomorphic to KB2_4, a positive triangle plus a fourth vertex negatively connected to all three. The paper further proves that selecting a size-K subset

What carries the argument

The central machinery is the time-series of complete signed graphs G_t^s(mu, R^N): for each trading day t, the edge between assets i and j is positive if (R_i^t - mu_i)(R_j^t - mu_j) >= 0 and negative otherwise. This graph is always balanced, meaning every triangle is of type T0 or T2. Two derived objects carry the argument: the hedge score h(n,T), the fraction of days on which asset n moves opposite to another asset relative to their means, and the labeled 4-clique KB2_4—a positive triangle with a fourth vertex negatively connected to all of its vertices. The proof chain claims balanced triangles encode positive skewness contributions and KB2_4 encodes negative kurtosis contributions, conve

Load-bearing premise

The load-bearing premise is that maximizing the signed-motif counts (balanced triangles and KB2_4 cliques) genuinely drives the full skewness and kurtosis of the portfolio, even though the true moments also contain repeated-index terms whose signs are controlled by individual asset behavior rather than by triangle or clique geometry.

What would settle it

Compute, for a fixed window of daily returns, the difference between the realized portfolio skewness and the contribution attributable to distinct-index balanced triangles, and similarly for kurtosis versus KB2_4 cliques; if there exist two equal-size asset subsets where the subset with strictly fewer balanced triangles or KB2_4 cliques has higher realized skewness and lower kurtosis, then the claimed combinatorial encoding is not the dominant determinant of the moments.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the moment-to-motif dictionary is correct, portfolio construction can screen assets by a combinatorial score proportional to hedge-weighted mean return—an O(N) per-day procedure that does not require covariance estimation.
  • The NP-hardness of the KB2_4-densest subgraph problem means any exact higher-moment-aware dimensionality reduction following the kurtosis prescription is intractable, motivating approximation algorithms or heuristics.
  • The backtesting suggests that reduced universes of 20–50 stocks formed this way can match or exceed the annual return of the full 199-asset universe in most out-of-sample years, making the method especially attractive for equally weighted portfolios.
  • The reported insensitivity of results to K (20, 30, 40, 50) suggests the hedge score is selecting a stable core of assets rather than a knife-edge optimum.
  • If the interpretation holds, the framework gives a graph-theoretic justification for the finance practice of diversifying with negatively correlated assets.

Where Pith is reading between the lines

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

  • The repeated-index terms in the skewness and kurtosis expansions are explicitly set aside; a testable extension would be to check empirically whether the asset subsets selected by hedge score also happen to be the sets with high balanced-triangle and KB2_4 density, or whether performance gains come entirely from variance reduction through negative edges.
  • The balancedness theorem relies on using the unconditional mean over a fixed window to define deviations; recomputing deviations with rolling or conditional means would break the proof and would be a natural stress test for the higher-moment interpretation.
  • The framework could transfer to other high-dimensional selection tasks where higher-order interactions matter, such as feature selection in models with multi-way interactions, though the paper does not explore this connection.
  • The proposed hedge score is a simple unnormalized frequency count; a normalized variant that accounts for the magnitude of deviations might yield even cleaner separation between reduced and full universes.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces a time-indexed complete signed graph on assets whose edge sign is the sign of the product of deviations from daily mean returns. It defines a hedge score equal to average negative degree over the window and proposes selecting top-K assets by hedge-score-weighted expected return (OPT1, Eq. 6) as dimensionality reduction for Markowitz and equal-weight portfolios. It further claims a combinatorial interpretation of higher moments: maximizing balanced triangles corresponds to maximizing skewness and maximizing balanced 4-cliques of type KB2_4 corresponds to minimizing kurtosis. This motivates OPT2 (Eq. 7) and an NP-hardness proof for KB2_4-density maximization. Backtesting on 199 S&P 500 stocks (2006–2021) compares reduced-universe portfolios with full-universe portfolios on return, volatility, and Sharpe ratio.

Significance. Strengths: the signed graph construction is simple and fully specified; the hedge-score definition and Algorithm 1 are transparent and reproducible; K is the sole user-chosen parameter and no backtest overfitting is apparent; the empirical protocol is easy to follow. If the moment correspondence were true, the paper would connect structural balance to higher-moment portfolio theory and would open a new motif-density problem in signed graphs. However, the central correspondence is wrong: every nonzero triangle in the model is balanced, so the balanced-triangle objective is constant on every subset; repeated-index terms are not modeled; Theorem 3.4 fails under zero deviations; and the NP-hardness proof has a concrete gap. The empirical results are mixed (higher volatility except 2009, mixed Sharpe), so the abstract's "effectiveness" claim is not supported by clean evidence. Thus the theoretical contribution collapses even if the hedge-score heuristic may have independent merit.

major comments (4)
  1. [§3.2, Case III (Eq. 4)] The asserted sign correspondence is false. Let s_i = sign(R_i^t−μ_i). Edge signs are s_i s_j, so for a triangle the product of the three edge signs is (s_i s_j)(s_j s_k)(s_k s_i)=+1; every nonzero-deviation triangle is balanced (T0 or T2). But S^t_ijk = s_i s_j s_k is negative whenever exactly one or three deviations are negative. For s=(+,+,−) the triangle is T2 yet S<0; for s=(−,−,−) it is T0 yet S<0. Consequently the count of balanced triangles is C(K,3) for every K-subset and cannot encode skewness or drive selection; the abstract's main claim and OPT2's triangle term rest on this false premise.
  2. [Theorem 3.4] The proof assumes strict nonzero deviations, but Definition 3.3 assigns a positive edge to any product ≥0. If δ_i=0, δ_j=1, δ_k=−1, then edges (i,j) and (i,k) are positive and (j,k) is negative, so the triangle is T1 and G^s_t is not balanced. The theorem is false as stated; it holds only under an added nonzero-deviation assumption. This matters because the paper later uses the 'always balanced' property to trivialize the balanced-triangle objective.
  3. [§4.2, Theorem 4.3] The converse direction of the reduction is invalid because KB2_4 patterns not containing the auxiliary vertex l are ignored. Example with c=6: let H be K_5 plus an isolated vertex v. In the constructed signed graph, the set S=V_H∪{l} contains C(5,3)=10 KB2_4 patterns from l with each triangle of K_5 and another 10 from v with those same triangles, for a total of 20=C(6,3), although H has no clique of size 6. Thus the threshold argument for CLIQUE fails and NP-hardness is not established.
  4. [§5] The backtest validates only OPT1/Algorithm 1; OPT2, the higher-moment combinatorial objective, is never solved. Since OPT2 is the proposed vehicle for the higher-moment interpretation, the numerical experiments cannot support the abstract's claim that the framework—including higher moments—is validated. Moreover, Figures 4–5 as described show that reduced universes are more volatile except in 2009 and have mixed Sharpe ratios, so 'demonstrating effectiveness' overstates the evidence.
minor comments (5)
  1. [Eq. (6)] The objective is written twice with and without an explicit sum; use consistent notation.
  2. [§5.1] Ticker strings contain apparent typos such as 'A VY', 'F AST', 'W A T', and 'TR V'; these should be cleaned.
  3. [§3.2] The sentence that a desirable portfolio should favor balanced triangles 'since G^s_t is a balanced complete graph' is self-defeating: if every triangle is balanced, the objective is constant on every subset and selects nothing.
  4. [§5.1] The Sharpe ratio is defined as annual return divided by annual volatility; state whether this is intentional rather than the usual mean/std with a risk-free rate, and provide confidence intervals or significance tests for the reported differences.
  5. [Figures 3–5] The figures are described only verbally; quantitative comparisons should be reported in tables with effect sizes or standard errors so the reader can verify the 'moderately better' claim.

Circularity Check

2 steps flagged

Skewness/balanced-triangle correspondence is a constant by construction: every triangle is balanced because edge signs are defined from the same deviations, so the 'maximize balanced triangles' target has no content.

specific steps
  1. self definitional [§3, Definition 3.3; Theorem 3.4; §3.2, Case III; OPT2 (Eq. 7)]
    "Case III: All three indices are distinct. Then the sign of S^t_{ijk} is positive precisely when the triangle induced by vertices i, j, k in the signed graph G_s^t (μ, R^N) is balanced, i.e., when it is of type T0 or T2. ... These observations indicate that a desirable portfolio should favor asset subsets that include balanced triangles ... thereby promoting higher skewness, which is obviously true since G_s^t (μ, R^N) is a balanced complete graph."

    Under Definition 3.3, edge signs are pairwise products of the individual deviations: σ_ij = s_i s_j, where s_i = sign(R^t_i − μ_i). Hence for any triangle, σ_ij σ_jk σ_ik = (s_i s_j)(s_j s_k)(s_k s_i) = +1, so every triangle is balanced (T0 or T2) and the graph is balanced for every t (Theorem 3.4). Therefore any K-vertex subset contains exactly C(K,3) balanced triangles; the 'maximize balanced triangles' objective is a constant and cannot select assets or encode skewness. The Case III assertion is also not an implication of balance: deviations (+,+,-) form a balanced T2 triangle while S^t_{ijk} = s_i s_j s_k = −1. The claimed skewness/balanced-triangle correspondence is thus a relabeling of the edge-sign definition, not a derived prediction, and the paper itself says so ('obviously true s

  2. other [§3.2, kurtosis cases; OPT2 (Eq. 7); §5 empirical analysis]
    "These observations indicate that a desirable portfolio should favor asset subsets that include balanced triangles in the time series of signed graphs G_s^t (μ, R^N), thereby promoting higher skewness, which is obviously true since G_s^t (μ, R^N) is a balanced complete graph. On the other hand, kurtosis reduction is associated with the presence of negative edges and a large number of balance 4-cliques of type K^{B2}_4."

    The kurtosis side is not tested through the reported backtest in the way it is advertised. Section 5 explicitly says that because OPT2 is computationally hard, the empirical study uses only Algorithm 1 (hedge score screening), not the K^{B2}_4 objective. The claim that combinatorial motif counts drive the backtest is therefore an interpretation imported from the construction, not an output of the fitting procedure. This is a secondary definitional gap rather than a fitted-input prediction: the empirical hedge-score pipeline itself is not circular, since hedge scores are computed from the training window only and K is user-chosen.

full rationale

The paper's own derivation makes the skewness/balanced-triangle correspondence trivial by construction. Definition 3.3 assigns each edge sign from (R^t_i−μ_i)(R^t_j−μ_j), so the product of the three edge signs in any triangle is (+1); Theorem 3.4 then proves that every G_s^t is balanced. Consequently every K-vertex induced subgraph has exactly C(K,3) balanced triangles, and the balanced-triangle term in the claimed dimensionality-reduction objective carries no information. The separate Case III claim that S^t_{ijk} > 0 iff the triangle is balanced is also false (e.g. signs (+,+,-) produce a balanced T2 triangle with negative S^t_{ijk}); although that is a correctness defect rather than circularity, it shows the proposed 'combinatorial interpretation' is not an independent translation of the higher-moment algebra. The paper itself concedes the balance conclusion is 'obviously true since G_s^t ... is a balanced complete graph.' The empirical selection pipeline is not circular: hedge scores are evaluated on the training year, K ∈ {20,30,40,50} is fixed by the user, no parameters are fitted to the out-of-sample outcomes, and the reduced universes are evaluated in the following year. The self-citation for Theorem 3.2 ([34]) is an elementary inequality that is independently checkable and not load-bearing. The NP-hardness reduction for K^{B2}_4 is self-contained rather than cited, and the backtest does not tune its result on the test years. Overall, the central higher-moment claim is partially circular/empty by construction, so the score is 6 rather than 0–2; but the empirical heuristic retains independent content, which keeps it below 8–10.

Axiom & Free-Parameter Ledger

1 free parameters · 4 axioms · 0 invented entities

The only tunable quantity is the subset cardinality K, set to 20/30/40/50 by hand and labeled ad hoc in Section 5. The main axioms are the sign-binarization of daily returns, the reduction of higher moments to signed-motif counts, and the standard Markowitz/Sharpe evaluation criteria. No new physical entities are introduced.

free parameters (1)
  • K (reduced universe cardinality) = 20, 30, 40, 50
    Cardinality constraint in OPT1/OPT2; the paper states 'the choice of K lies with the user' and is 'inherently ad hoc' (Section 5). Tested values are not derived from any economic or statistical criterion.
axioms (4)
  • domain assumption Daily co-movement is binarized: sign((R_i^t - μ_i)(R_j^t - μ_j)), with zero treated as positive edge (Definition 3.3).
    This modeling choice discards magnitude and misclassifies zero-deviation pairs as positive edges; it makes Theorem 3.4 false when a return equals its mean.
  • ad hoc to paper Skewness is governed by balanced triangles alone (Section 3.2, Case III), with repeated-index terms assumed non-problematic.
    Section 3.2 itself shows terms with i=j have sign determined by a single asset's deviation; the paper does not bound these, so count of balanced triangles does not determine sign of S(w).
  • domain assumption Minimizing kurtosis is equivalent to maximizing the count of KB2_4 4-cliques (Section 3.2, Case II).
    A KB2_4 pattern indeed has negative product of four deviations, but kurtosis is a magnitude-weighted sum over all ordered index tuples; count-maximization ignores magnitudes and repeated-index terms.
  • standard math Markowitz mean-variance optimization and Sharpe maximization are the correct performance criteria (Section 2, Eq. 1-2 and Section 5.1, Eq. 8).
    Both are standard tools in the cited literature; the paper does not need to derive them, but their validity is assumed for the backtest conclusions.

reviewed 2026-08-02 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Signed network models for dimensionality reduction of portfolio optimization." pith.science (2026). https://pith.science/paper/CTHCC4BM

@misc{pith2026260221362,
  author       = {Pith},
  title        = {Pith review of: Signed network models for dimensionality reduction of portfolio optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CTHCC4BM}},
  note         = {Machine review of arXiv:2602.21362}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

In this paper, we develop a time-series-based signed network model for dimensionality reduction in portfolio optimization, grounded in Markowitz's portfolio theory and extended to incorporate higher-order moments of asset return distributions. Unlike traditional correlation-based approaches, we construct a complete signed graph for each trading day within a specified time window, where the sign of an edge between a pair of assets is determined by the relative behavior of their log returns with respect to their mean returns. Within this framework, we introduce a combinatorial interpretation of higher-order moments, showing that maximizing skewness and minimizing kurtosis correspond to maximizing balanced triangles and balanced 4-cliques with specific signed edge configurations respectively. We establish that the latter leads to an NP-hard combinatorial optimization problem, while the former is naturally guaranteed by the structural properties of the signed graph model. Based on this interpretation, we propose a dimensionality reduction method using a combinatorial formulation of the mean-variance optimization problem through a combinatorial hedge score metric for assets. The proposed framework is validated through extensive backtesting on 199 S\&P 500 assets over a 16-year period (2006 - 2021), demonstrating the effectiveness of reduced asset universes for portfolio construction using both Markowitz optimization and equally weighted strategy.

Figures

Figures reproduced from arXiv: 2602.21362 by Bibhas Adhikari.

Figure 1
Figure 1. Figure 1: (a) Threshold function [46] for signed network formation. [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The possible 4-cliques in Gs t (µ, RN ). Edges in green and red indicate positive and negative edges respectively. 4 Dimensionality reduction of portfolio optimization In this section, we develop a dimensionality reduction framework based on the proposed hedge￾score formulation derived from Markowitz’s portfolio optimization model, and further extend it by incorporating higher-order moments into the portfo… view at source ↗
Figure 3
Figure 3. Figure 3: Annual Return: (a) K = 20, (b) K = 30, (c) K = 40, (d) K = 50. For backtesting on the out-of-sample data, we consider three metrics: annual return, annual volatility, and Sharpe value for comparing the portfolios. The daily log returns for each asset j at a local time t (a day) is computed as rj,t = log(Pj,t/Pj,t−1), where Pi,t is the price of asset j at time t. Then for a given weight vector w, the daily … view at source ↗
Figure 4
Figure 4. Figure 4: Annual Volatility: (a) K = 20, (b) K = 30, (c) K = 40, (d) K = 50. We would like to mention that the selection of the reduced universe size K in the proposed 14 [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Sharpe ratio: (a) K = 20, (b) K = 30, (c) K = 40, (d) K = 50. framework is inherently ad hoc since the choice of K lies with the user. Although the chosen values yield stable and interpretable empirical results, they are not derived from an explicit economic ob￾jective or statistical criterion. Determining optimal or adaptive choices of K potentially as functions of market conditions, asset universe size, … view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

59 extracted references · 2 linked inside Pith

  1. [1]

    Portfolio selection, the journal of finance

    Harry M Markowitz. Portfolio selection, the journal of finance. 7 (1).N, 1:71–91, 1952

    1952

  2. [2]

    Optimal portfolio allocation under higher moments

    Eric Jondeau and Michael Rockinger. Optimal portfolio allocation under higher moments. European Financial Management, 12(1):29–55, 2006

    2006

  3. [3]

    Portfolio selection with higher moments.Quantitative Finance, 10(5):469–485, 2010

    Campbell R Harvey, John C Liechty, Merrill W Liechty, and Peter M¨ uller. Portfolio selection with higher moments.Quantitative Finance, 10(5):469–485, 2010

    2010

  4. [4]

    Machine learning and portfolio optimization.Management Science, 64(3):1136–1154, 2018

    Gah-Yi Ban, Noureddine El Karoui, and Andrew EB Lim. Machine learning and portfolio optimization.Management Science, 64(3):1136–1154, 2018

    2018

  5. [5]

    Portfolio management with higher mo- ments: the cardinality impact.International Transactions in Operational Research, 26(6):2531– 2560, 2019

    Rui Pedro Brito, H´ elder Sebasti˜ ao, and Pedro Godinho. Portfolio management with higher mo- ments: the cardinality impact.International Transactions in Operational Research, 26(6):2531– 2560, 2019. 18

    2019

  6. [6]

    Gainers and losers with higher order portfolio risk optimization.Physica A: Statistical Mechanics and its Applications, 563:125416, 2021

    Saira Ashfaq, Usman Ayub, Ghulam Mujtaba, Naveed Raza, and Saqib Gulzar. Gainers and losers with higher order portfolio risk optimization.Physica A: Statistical Mechanics and its Applications, 563:125416, 2021

    2021

  7. [7]

    Solving high-order portfolios via successive convex approxi- mation algorithms.IEEE Transactions on Signal Processing, 69:892–904, 2021

    Rui Zhou and Daniel P Palomar. Solving high-order portfolios via successive convex approxi- mation algorithms.IEEE Transactions on Signal Processing, 69:892–904, 2021

    2021

  8. [8]

    A theoretical generalization of the markowitz model incorporating skewness and kurtosis.Quantitative Finance, 23(5):877–886, 2023

    Pierpaolo Uberti. A theoretical generalization of the markowitz model incorporating skewness and kurtosis.Quantitative Finance, 23(5):877–886, 2023

    2023

  9. [9]

    Pgp for portfolio optimization: application to esg index family.Annals of Operations Research, 347(1):405–417, 2025

    Ilyes Abid, Christian Urom, Jonathan Peillex, Majdi Karmani, and Gideon Ndubuisi. Pgp for portfolio optimization: application to esg index family.Annals of Operations Research, 347(1):405–417, 2025

    2025

  10. [10]

    Some np-complete problems in quadratic and nonlinear programming

    Katta G Murty and Santosh N Kabadi. Some np-complete problems in quadratic and nonlinear programming. Technical report, 1985

    1985

  11. [11]

    Optimal versus naive diversification: How inefficient is the 1/n portfolio strategy?The review of Financial studies, 22(5):1915–1953, 2009

    Victor DeMiguel, Lorenzo Garlappi, and Raman Uppal. Optimal versus naive diversification: How inefficient is the 1/n portfolio strategy?The review of Financial studies, 22(5):1915–1953, 2009

    1915

  12. [12]

    The properties of equally weighted risk contribution portfolios.Journal of portfolio management, 36(4):60, 2010

    S´ ebastien Maillard, Thierry Roncalli, and J´ erˆ ome Te ¨ ıletche. The properties of equally weighted risk contribution portfolios.Journal of portfolio management, 36(4):60, 2010

    2010

  13. [13]

    The 1/n investment strategy is optimal under high model ambiguity.Journal of Banking & Finance, 36(2):410–417, 2012

    Georg Ch Pflug, Alois Pichler, and David Wozabal. The 1/n investment strategy is optimal under high model ambiguity.Journal of Banking & Finance, 36(2):410–417, 2012

    2012

  14. [14]

    The equally weighted portfolio still remains a challenging benchmark.International Economics, 179:100525, 2024

    Matteo Gelmini and Pierpaolo Uberti. The equally weighted portfolio still remains a challenging benchmark.International Economics, 179:100525, 2024

    2024

  15. [15]

    Combining the minimum-variance and equally- weighted portfolios: Can portfolio performance be improved?Economic Modelling, 80:260–274, 2019

    Chonghui Jiang, Jiangze Du, and Yunbi An. Combining the minimum-variance and equally- weighted portfolios: Can portfolio performance be improved?Economic Modelling, 80:260–274, 2019

    2019

  16. [16]

    A network perspective of the stock market.Journal of Empirical Finance, 17(4):659–667, 2010

    K Tse Chi, Jing Liu, and Francis CM Lau. A network perspective of the stock market.Journal of Empirical Finance, 17(4):659–667, 2010

    2010

  17. [17]

    Hierarchical structure in financial markets.The European Physical Journal B-Condensed Matter and Complex Systems, 11:193–197, 1999

    Rosario N Mantegna. Hierarchical structure in financial markets.The European Physical Journal B-Condensed Matter and Complex Systems, 11:193–197, 1999

    1999

  18. [18]

    Nets: Network estimation for time series.Journal of Applied Econometrics, 34(3):347–364, 2019

    Matteo Barigozzi and Christian Brownlees. Nets: Network estimation for time series.Journal of Applied Econometrics, 34(3):347–364, 2019

    2019

  19. [19]

    Correlation analysis and systemic risk measurement of regional, financial and global stock indices.Physica A: Statistical Mechanics and its Applications, 542:122653, 2020

    Lin Chen, Qian Han, Zhilin Qiao, and H Eugene Stanley. Correlation analysis and systemic risk measurement of regional, financial and global stock indices.Physica A: Statistical Mechanics and its Applications, 542:122653, 2020

    2020

  20. [20]

    Financial network systemic risk contributions.Review of Finance, 19(2):685–738, 2015

    Nikolaus Hautsch, Julia Schaumburg, and Melanie Schienle. Financial network systemic risk contributions.Review of Finance, 19(2):685–738, 2015

    2015

  21. [21]

    A network approach to portfolio selection.Journal of Empirical Finance, 38:157–180, 2016

    Gustavo Peralta and Abalfazl Zareei. A network approach to portfolio selection.Journal of Empirical Finance, 38:157–180, 2016. 19

    2016

  22. [22]

    Signature-based portfolio allocation: a network approach.Applied Network Science, 9(1):54, 2024

    Marco Gregnanin, Yanyi Zhang, Johannes De Smedt, Giorgio Gnecco, and Maurizio Parton. Signature-based portfolio allocation: a network approach.Applied Network Science, 9(1):54, 2024

    2024

  23. [23]

    Network science: a useful tool in economics and finance

    Dror Y Kenett and Shlomo Havlin. Network science: a useful tool in economics and finance. Mind & Society, 14:155–167, 2015

    2015

  24. [24]

    Networks in the understanding of economic behaviors.Journal of economic perspectives, 28(4):3–22, 2014

    Matthew O Jackson. Networks in the understanding of economic behaviors.Journal of economic perspectives, 28(4):3–22, 2014

    2014

  25. [25]

    Systemic risk and stability in financial networks.American Economic Review, 105(2):564–608, 2015

    Daron Acemoglu, Asuman Ozdaglar, and Alireza Tahbaz-Salehi. Systemic risk and stability in financial networks.American Economic Review, 105(2):564–608, 2015

    2015

  26. [26]

    A new method to estimate the noise in financial correlation matrices.Journal of Physics A: Mathematical and General, 36(12):3009, 2003

    Thomas Guhr and Bernd K¨ alber. A new method to estimate the noise in financial correlation matrices.Journal of Physics A: Mathematical and General, 36(12):3009, 2003

    2003

  27. [27]

    Noise dressing of financial correlation matrices.Physical review letters, 83(7):1467, 1999

    Laurent Laloux, Pierre Cizeau, Jean-Philippe Bouchaud, and Marc Potters. Noise dressing of financial correlation matrices.Physical review letters, 83(7):1467, 1999

    1999

  28. [28]

    Noisy covariance matrices and portfolio optimization ii.Physica A: Statistical Mechanics and its Applications, 319:487–494, 2003

    Szil´ ard Pafka and Imre Kondor. Noisy covariance matrices and portfolio optimization ii.Physica A: Statistical Mechanics and its Applications, 319:487–494, 2003

    2003

  29. [29]

    Noise sensitivity of portfolio selection under various risk measures.Journal of Banking & Finance, 31(5):1545–1573, 2007

    Imre Kondor, Szil´ ard Pafka, and G´ abor Nagy. Noise sensitivity of portfolio selection under various risk measures.Journal of Banking & Finance, 31(5):1545–1573, 2007

    2007

  30. [30]

    The effects of errors in means, variances, and correlations on the mean-variance framework.Quantitative Finance, 22(10):1893–1903, 2022

    Munki Chung, Yongjae Lee, Jang Ho Kim, Woo Chang Kim, and Frank J Fabozzi. The effects of errors in means, variances, and correlations on the mean-variance framework.Quantitative Finance, 22(10):1893–1903, 2022

    1903

  31. [31]

    Partial correlation financial networks.Applied Net- work Science, 5(1):11, 2020

    Tristan Millington and Mahesan Niranjan. Partial correlation financial networks.Applied Net- work Science, 5(1):11, 2020

    2020

  32. [32]

    Forecasting financial market structure from network features using machine learning

    Douglas Castilho, Tharsis TP Souza, Soong Moon Kang, Jo˜ ao Gama, and Andr´ e CPLF de Car- valho. Forecasting financial market structure from network features using machine learning. Knowledge and Information Systems, 66(8):4497–4525, 2024

    2024

  33. [33]

    Shrinkage estimation with reinforcement learning of large variance matrices for portfolio selection.Intelligent Systems with Applications, 17:200181, 2023

    Giulio Mattera and Raffaele Mattera. Shrinkage estimation with reinforcement learning of large variance matrices for portfolio selection.Intelligent Systems with Applications, 17:200181, 2023

    2023

  34. [34]

    Signed network models for portfolio optimization.arXiv preprint arXiv:2510.05377, 2025

    Bibhas Adhikari. Signed network models for portfolio optimization.arXiv preprint arXiv:2510.05377, 2025

    arXiv 2025

  35. [35]

    Portfolio optimization with digitized counterdiabatic quantum al- gorithms.Physical Review Research, 4(4):043204, 2022

    Narendra N Hegade, Pranav Chandarana, Koushik Paul, Xi Chen, Francisco Albarr´ an- Arriagada, and E Solano. Portfolio optimization with digitized counterdiabatic quantum al- gorithms.Physical Review Research, 4(4):043204, 2022

    2022

  36. [36]

    Train-and-scaling the quantum alternating operator ansatz to solve portfolio diversification

    Hannes Leipold and Sarvagya Upadhyay. Train-and-scaling the quantum alternating operator ansatz to solve portfolio diversification. In2024 IEEE International Conference on Quantum Computing and Engineering (QCE), volume 2, pages 132–137. IEEE, 2024

    2024

  37. [37]

    The lie algebra of xy-mixer topologies and warm starting qaoa for constrained optimization.arXiv preprint arXiv:2505.18396, 2025

    Steven Kordonowy and Hannes Leipold. The lie algebra of xy-mixer topologies and warm starting qaoa for constrained optimization.arXiv preprint arXiv:2505.18396, 2025. 20

    Pith/arXiv arXiv 2025

  38. [38]

    Scaling portfolio diversification with quantum circuit cutting techniques.arXiv preprint arXiv:2506.08947, 2025

    Vicente P Soloviev, Antonio M´ arquez Romero, Josh Kirsopp, and Michal Krompiec. Scaling portfolio diversification with quantum circuit cutting techniques.arXiv preprint arXiv:2506.08947, 2025

    Pith/arXiv arXiv 2025

  39. [39]

    Best practices for portfolio optimization by quantum computing, experimented on real quantum devices.Scientific Reports, 13(1):19434, 2023

    Giuseppe Buonaiuto, Francesco Gargiulo, Giuseppe De Pietro, Massimo Esposito, and Marco Pota. Best practices for portfolio optimization by quantum computing, experimented on real quantum devices.Scientific Reports, 13(1):19434, 2023

    2023

  40. [40]

    A mathematical bibliography of signed and gain graphs and allied areas

    Thomas Zaslavsky. A mathematical bibliography of signed and gain graphs and allied areas. The Electronic Journal of Combinatorics, pages DS8–Dec, 2012

    2012

  41. [41]

    Structural balance: a generalization of heider’s theory

    Dorwin Cartwright and Frank Harary. Structural balance: a generalization of heider’s theory. Psychological review, 63(5):277, 1956

    1956

  42. [42]

    On the notion of balance of a signed graph.Michigan Mathematical Journal, 2(2):143–146, 1953

    Frank Harary. On the notion of balance of a signed graph.Michigan Mathematical Journal, 2(2):143–146, 1953

    1953

  43. [43]

    Balance and frustration in signed networks.Journal of Complex Networks, 7(2):163–189, 2019

    Samin Aref and Mark C Wilson. Balance and frustration in signed networks.Journal of Complex Networks, 7(2):163–189, 2019

    2019

  44. [44]

    Measuring the balance of signed networks and its application to sign prediction.Journal of Statistical Mechanics: Theory and Experiment, 2017(6):063302, 2017

    Ranveer Singh and Bibhas Adhikari. Measuring the balance of signed networks and its application to sign prediction.Journal of Statistical Mechanics: Theory and Experiment, 2017(6):063302, 2017

    2017

  45. [45]

    PhD thesis, Institute of Cross-Disciplinary Physics and Complex Systems, IFISC, 2025

    Fernando Diaz-Diaz.Mathematical analysis of signed networks: structure and dynamics. PhD thesis, Institute of Cross-Disciplinary Physics and Complex Systems, IFISC, 2025

    2025

  46. [46]

    Signed graphs for portfolio analysis in risk management.IMA Journal of management mathematics, 13(3):201–210, 2002

    Frank Harary, Meng-Hiot Lim, and Donald C Wunsch. Signed graphs for portfolio analysis in risk management.IMA Journal of management mathematics, 13(3):201–210, 2002

    2002

  47. [47]

    Separator-based data reduction for signed graph balancing.Journal of combinatorial optimization, 20(4):335–360, 2010

    Falk H¨ uffner, Nadja Betzler, and Rolf Niedermeier. Separator-based data reduction for signed graph balancing.Journal of combinatorial optimization, 20(4):335–360, 2010

    2010

  48. [48]

    The structure of stock markets as signed networks.Journal of Industrial and Systems Engineering, 13(1):136–146, 2020

    Maryam Ehsani. The structure of stock markets as signed networks.Journal of Industrial and Systems Engineering, 13(1):136–146, 2020

    2020

  49. [49]

    The maximum balanced subgraph of a signed graph: Appli- cations and solution approaches.European Journal of Operational Research, 236(2):473–487, 2014

    Rosa Figueiredo and Yuri Frota. The maximum balanced subgraph of a signed graph: Appli- cations and solution approaches.European Journal of Operational Research, 236(2):473–487, 2014

    2014

  50. [50]

    Applications of signed graphs to portfolio turnover analysis.Procedia-Social and Behavioral Sciences, 211:1203–1209, 2015

    B Vasanthi, S Arumugam, Atulya K Nagar, and Sovan Mitra. Applications of signed graphs to portfolio turnover analysis.Procedia-Social and Behavioral Sciences, 211:1203–1209, 2015

    2015

  51. [51]

    Global balance and systemic risk in financial correlation networks.Physica A: Statistical Mechanics and its Applications, page 130698, 2025

    Paolo Bartesaghi, Fernando Diaz-Diaz, Rosanna Grassi, and Pierpaolo Uberti. Global balance and systemic risk in financial correlation networks.Physica A: Statistical Mechanics and its Applications, page 130698, 2025

    2025

  52. [52]

    Introduction to corre- lation networks: Interdisciplinary approaches beyond thresholding.Physics Reports, 1136:1–39, 2025

    Naoki Masuda, Zachary M Boyd, Diego Garlaschelli, and Peter J Mucha. Introduction to corre- lation networks: Interdisciplinary approaches beyond thresholding.Physics Reports, 1136:1–39, 2025. 21

    2025

  53. [53]

    A survey on gaps between mean-variance approach and exponential growth rate approach for portfolio optimization.ACM Computing Surveys (CSUR), 55(2):1–36, 2022

    Zhao-Rong Lai and Haisheng Yang. A survey on gaps between mean-variance approach and exponential growth rate approach for portfolio optimization.ACM Computing Surveys (CSUR), 55(2):1–36, 2022

    2022

  54. [54]

    Cambridge university press, 1999

    Rosario N Mantegna and H Eugene Stanley.Introduction to econophysics: correlations and complexity in finance. Cambridge university press, 1999

    1999

  55. [55]

    Generalized autoregressive conditional heteroskedasticity.Journal of economet- rics, 31(3):307–327, 1986

    Tim Bollerslev. Generalized autoregressive conditional heteroskedasticity.Journal of economet- rics, 31(3):307–327, 1986

    1986

  56. [56]

    Autoregressive conditional heteroscedasticity with estimates of the variance of united kingdom inflation.Econometrica: Journal of the econometric society, pages 987–1007, 1982

    Robert F Engle. Autoregressive conditional heteroscedasticity with estimates of the variance of united kingdom inflation.Econometrica: Journal of the econometric society, pages 987–1007, 1982

    1982

  57. [57]

    Is gold a hedge or a safe haven? an analysis of stocks, bonds and gold.Financial review, 45(2):217–229, 2010

    Dirk G Baur and Brian M Lucey. Is gold a hedge or a safe haven? an analysis of stocks, bonds and gold.Financial review, 45(2):217–229, 2010

    2010

  58. [58]

    The k-clique densest subgraph problem

    Charalampos Tsourakakis. The k-clique densest subgraph problem. InProceedings of the 24th international conference on world wide web, pages 1122–1132, 2015

    2015

  59. [59]

    Reducibility among combinatorial problems

    Richard M Karp. Reducibility among combinatorial problems. In50 Years of Integer Pro- gramming 1958-2008: from the Early Years to the State-of-the-Art, pages 219–241. Springer, 2009. 22

    1958

This paper was first reviewed by deepseek-v4-flash on August 2, 2026.