Paper: arXiv 2308.04769

Authors: Ryo Hidaka, Yohei Hamakawa, Jun Nakayama, Kosuke Tatsumura

Abstract

Correlation-diversified portfolios can be constructed by finding the maximum independent sets (MISs) in market graphs with edges corresponding to correlations between two stocks. The computational complexity to find the MIS increases exponentially as the size of the market graph increases, making the MIS selection in a large-scale market graph difficult. Here we construct a diversified portfolio by solving the MIS problem for a large-scale market graph with a combinatorial optimization solver (an Ising machine) based on a quantum-inspired algorithm called simulated bifurcation (SB) and investigate the investment performance of the constructed portfolio using long-term historical market data. Comparisons using stock universes of various sizes [“TOPIX 100, Nikkei 225, TOPIX 1000, and TOPIX (including approximately 2,000 constituents)”] show that the SB-based solver outperforms conventional MIS solvers in terms of computation-time and solution-accuracy. By using the SB-based solver, we optimized the parameters of a MIS portfolio strategy through iteration of the backcast simulation that calculates the performance of the MIS portfolio strategy based on a large-scale universe covering more than 1,700 Japanese stocks for a long period of 10 years. It has been found that the best MIS portfolio strategy (Sharpe ratio = 1.16, annualized return/risk = 16.3%/14.0%) outperforms the major indices such as TOPIX (0.66, 10.0%/15.2%) and MSCI Japan Minimum Volatility Index (0.64, 7.7%/12.1%) for the period from 2013 to 2023.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced combinatorial optimization and graph theory, mapping the MIS problem to an Ising model and utilizing a quantum-inspired simulated bifurcation algorithm, which is mathematically dense. It demonstrates high empirical rigor by implementing an FPGA-based solver, running extensive backcasts over 10 years on a large dataset (1,700+ stocks), and reporting detailed performance metrics like Sharpe ratios against benchmarks.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Construct correlation-diversified portfolios<br>for large-scale markets"] --> B["Method: Maximum Independent Set<br>Find MIS in market graph with<br>correlation edges using Ising machine"]
  
  B --> C["Data: Long-term historical data<br>Japanese stocks (1,700+ stocks)<br>10-year period (2013-2023)"]
  
  C --> D["Computation: Simulated Bifurcation<br>SB-based solver<br>vs conventional solvers"]
  
  D --> E["Parameter Optimization<br>Iterative backcast simulation<br>for MIS portfolio strategy"]
  
  E --> F["Key Findings: SB solver outperforms<br>Sharpe ratio 1.16 (16.3%/14.0%)<br>TOPIX 0.66, MSCI MinVol 0.64"]
  
  F --> G["Outcomes: Large-scale market<br>feasible MIS portfolio construction<br>via quantum-inspired computing"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a simulated bifurcation (SB) based FPGA accelerator to solve the maximum independent set (MIS) problem for constructing correlation‑diversified portfolios.

Source sentence
We have developed a look-aside type combinatorial optimization accelerator (an Ising machine) based on a quantum‑inspired parallelizable algorithm called simulated bifurcation (SB), which enables the long‑term backcast simulation of a diversified portfolio strategy based on the selection of the maximum independent set

Assumes 6

  • Portfolio is rebalanced monthly using closing prices and dividends. execution
    Source sentence
    The portfolio is rebalanced monthly; the portfolio selection and weighting are carried out every month (at the end of each month).
  • Trading costs are 0.1% of total transaction amount. execution
    Source sentence
    we take into account the trading costs corresponding to 0.1% of the total amount of transactions (the sum of the amount of shares purchased and the absolute amount of shares sold).
  • Correlation threshold θ is fixed for the whole simulation (not changed monthly). model
    Source sentence
    we determine the best θ by analyzing the long‑term performance of the MIS portfolio strategy for one of various θ and then assume using a fixed θ for the MIS strategy (not changed monthly).
  • Correlation matrix is computed from the last three years of price data each month. data
    Source sentence
    At the beginning of the monthly processing, the correlation matrix C and the volatility {vi} are calculated based on the price data for the last three years before the month as described in Sec. 2.2.1 and Sec. 2.1.2.
Show 2 more
  • Universe consists of stocks that remained listed for the analysis period. data
    Source sentence
    From the candidates, the stocks that have remained listed for the analysis period consist of the stock universe.
  • Dividends are included in returns for the MIS strategy. execution
    Source sentence
    the dividend (if exists) is considered.

Data

Backtest sample: 2013 – 2023 · Japanese stocks · equities

  • Japanese stock price data (TOPIX constituents) · Tokyo Stock Exchange market-prices public
    Apr. 1, 2013 to Mar. 31, 2023 · 1,747 stocks corresponding to TOPIX
    Source sentence
    By using the SBM‑based portfolio simulator described in Sec. 3.1, we simulate the performance of the MIS portfolio strategy over 10 years (Apr. 1, 2013, to Mar. 31, 2023) for a large universe including 1,747 stocks corresponding to TOPIX.

Costs

Modeled Costs are part of the reported headline results.

0.1% of total transaction amount (sum of shares bought and sold) is applied to the MIS strategy results.

Source sentence
we take into account the trading costs corresponding to 0.1% of the total amount of transactions (the sum of the amount of shares purchased and the absolute amount of shares sold).

Finds 5

Reported performance (MIS portfolio strategy): Sharpe 1.16 (benchmark 0.66) · annual return 16.3%

  • The best MIS portfolio (θ=0.23, inverse‑volatility weighting) achieved Sharpe ratio 1.16, annualized return 16.3% and risk 14.0%, outperforming TOPIX and MSCI Japan Minimum Volatility Index. performance
    Source sentence
    It has been found that the best MIS portfolio strategy (Sharpe ratio = 1.16, annualized return/risk = 16.3%/14.0%) outperforms the major indices such as TOPIX (0.66, 10.0%/15.2%) and MSCI Japan Minimum Volatility Index (0.64, 7.7%/12.1%) for the period from 2013 to 2023.
  • The SB‑based MIS solver is 6,230 times faster than a conventional heuristic solver on a 2,026‑node market graph. performance
    Source sentence
    At a large‑scale market graph of 2,026 nodes (too large for the exact solver to handle), the SB‑based solver is 6,230 times faster than the conventional heuristic solver.
  • SB‑based solver provides solutions about 20% larger than those from a conventional heuristic (NetworkX). performance
    Source sentence
    The SB‑based MIS solver provides a good approximate solution very close to the exact solution in a much shorter time than an exact‑solution solver (OR‑tools) and also finds the independent sets with a remarkably larger size (roughly 20%) than a conventional heuristic solver (NetworkX).
  • Increasing the correlation threshold θ reduces risk and increases portfolio size, with Sharpe ratio peaking around θ=0.18‑0.19. performance
    Source sentence
    As the correlation threshold θ increases, the number of excluded edges increases, and therefore the edges get more sparse, as a result, the size of independent set increases... The
  • The MIS strategy’s cumulative excess return over TOPIX was especially high (49.0% annualized) during Sep. 2016‑Dec. 2017. performance
    Source sentence
    The annualized return of the MIS strategy for the period from Sep. 2016 and Dec. 2017 is 49.0%, remarkably larger than that of TOPIX (22.7%).

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 53 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.