Roy H. Kwon
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Roy H. Kwon has a total of 65 co-authors. Here are the top co-authors by number of shared publications:
- David Islip
Department of Mechanical and Industrial Engineering, University of Toronto, Toronto, Canada
7 shared publications - Giorgio Costa
Millennium Management, New York, NY, 10022, USA
4 shared publications Andrew Butler
Department of Mechanical and Industrial Engineering, University of Toronto, 5 King's College Rd, Toronto, ON, M5S 3G8, Canada
3 shared publications- Hassan T. Anis3 shared publications
- Oleksandr Romanko
SS&C Algorithmics, Toronto, Canada
3 shared publications - Yuri Lawryshyn
University of Toronto
3 shared publications
Latest Publications
- 10 November 2026
- Published by Springer Nature in Computational Optimization and Applications
- p. 1-43
Contextual scenario generation for two-stage stochastic programming
- 10 November 2026
- Published by Springer Nature in Computational Optimization and Applications
- p. 1-43
Contextual scenario generation for two-stage stochastic programming
- 15 August 2026
- Published by Springer Nature in Operations Research Forum
- Vol. 7 (3) , 97
Generative AI-Enhanced Sector-Based Investment Portfolio Construction
- 15 August 2026
- Published by Springer Nature in Operations Research Forum
- Vol. 7 (3) , 97
Generative AI-Enhanced Sector-Based Investment Portfolio Construction
- 30 June 2026
- Published by Springer Nature in Journal of Optimization Theory and Applications
- Vol. 210 (1) , 16
Mixed-Integer Optimization for Cross-Validated Feature Selection in Linear Classification
- 30 June 2026
- Published by Springer Nature in Journal of Optimization Theory and Applications
- Vol. 210 (1) , 16
Mixed-Integer Optimization for Cross-Validated Feature Selection in Linear Classification
- 30 April 2026
- Published by Springer Nature in Annals of Operations Research
- Vol. 366 (1) , 169-200
A regime-switching decision support system for dynamic portfolio optimization
- 30 April 2026
- Published by Springer Nature in Annals of Operations Research
- Vol. 366 (1) , 169-200
A regime-switching decision support system for dynamic portfolio optimization
- 25 February 2026
- Published by Springer Nature in Optimization and Engineering
- p. 1-23
Smart classify and optimize for portfolio construction - a differentiable quadratic programming approach
- 25 February 2026
- Published by Springer Nature in Optimization and Engineering
- p. 1-23
Smart classify and optimize for portfolio construction - a differentiable quadratic programming approach
- Open Access
Generative AI-enhanced Sector-based Investment Portfolio Construction
This paper investigates how Large Language Models (LLMs) from leading providers (OpenAI, Google, Anthropic, DeepSeek, and xAI) can be applied to quantitative sector-based portfolio construction. We use LLMs to identify investable universes of stocks within S&P 500 sector indices and evaluate how their selections perform when combined with classical portfolio optimization methods. Each model was prompted to select and weight 20 stocks per sector, and the resulting portfolios were compared with their respective sector indices across two distinct out-of-sample periods: a stable market phase (January-March 2025) and a volatile phase (April-June 2025). Our results reveal a strong temporal dependence in LLM portfolio performance. During stable market conditions, LLM-weighted portfolios frequently outperformed sector indices on both cumulative return and risk-adjusted (Sharpe ratio) measures. However, during the volatile period, many LLM portfolios underperformed, suggesting that current models may struggle to adapt to regime shifts or high-volatility environments underrepresented in their training data. Importantly, when LLM-based stock selection is combined with traditional optimization techniques, portfolio outcomes improve in both performance and consistency. This study contributes one of the first multi-model, cross-provider evaluations of generative AI algorithms in investment management. It highlights that while LLMs can effectively complement quantitative finance by enhancing stock selection and interpretability, their reliability remains market-dependent. The findings underscore the potential of hybrid AI-quantitative frameworks, integrating LLM reasoning with established optimization techniques, to produce more robust and adaptive investment strategies.
Open AccessGenerative AI-enhanced Sector-based Investment Portfolio Construction
This paper investigates how Large Language Models (LLMs) from leading providers (OpenAI, Google, Anthropic, DeepSeek, and xAI) can be applied to quantitative sector-based portfolio construction. We use LLMs to identify investable universes of stocks within S&P 500 sector indices and evaluate how their selections perform when combined with classical portfolio optimization methods. Each model was prompted to select and weight 20 stocks per sector, and the resulting portfolios were compared with their respective sector indices across two distinct out-of-sample periods: a stable market phase (January-March 2025) and a volatile phase (April-June 2025). Our results reveal a strong temporal dependence in LLM portfolio performance. During stable market conditions, LLM-weighted portfolios frequently outperformed sector indices on both cumulative return and risk-adjusted (Sharpe ratio) measures. However, during the volatile period, many LLM portfolios underperformed, suggesting that current models may struggle to adapt to regime shifts or high-volatility environments underrepresented in their training data. Importantly, when LLM-based stock selection is combined with traditional optimization techniques, portfolio outcomes improve in both performance and consistency. This study contributes one of the first multi-model, cross-provider evaluations of generative AI algorithms in investment management. It highlights that while LLMs can effectively complement quantitative finance by enhancing stock selection and interpretability, their reliability remains market-dependent. The findings underscore the potential of hybrid AI-quantitative frameworks, integrating LLM reasoning with established optimization techniques, to produce more robust and adaptive investment strategies.
- Open Access
- 16 September 2025
- Published by Springer Nature in Research Square
Smart classify and optimize for portfolio construction - a differentiable quadratic programming approach
Portfolio construction requires not only allocating capital but also selecting which assets to include. In practice, these two steps are typically separated: a clas-sifier such as a support vector machine (SVM) screens assets, after which a mean–variance optimizer (MVO) allocates capital on the retained set. This two-stage design enforces a fixed screening boundary, ignoring the fact that investors with different return targets may prefer different sets of assets. We propose an end-to-end framework that integrates screening and allocation into a single dif-ferentiable optimization pipeline. Both the SVM and the MVO are represented as quadratic programming (QP) layers, connected through a differentiable gat-ing mechanism that enables gradient-based training. This formulation adapts the screening hyperplane to the investor’s return goal, effectively coupling asset selection with allocation. Empirical results on sector-diversified U.S. equities demonstrate that the integrated model delivers improved risk–return trade-offs over the two-stage baseline, with pronounced gains in volatile market regimes, representing the first differentiable integration of SVM and MVO.
- 16 September 2025
- Published by Springer Nature in Research Square
Open AccessSmart classify and optimize for portfolio construction - a differentiable quadratic programming approach
Portfolio construction requires not only allocating capital but also selecting which assets to include. In practice, these two steps are typically separated: a clas-sifier such as a support vector machine (SVM) screens assets, after which a mean–variance optimizer (MVO) allocates capital on the retained set. This two-stage design enforces a fixed screening boundary, ignoring the fact that investors with different return targets may prefer different sets of assets. We propose an end-to-end framework that integrates screening and allocation into a single dif-ferentiable optimization pipeline. Both the SVM and the MVO are represented as quadratic programming (QP) layers, connected through a differentiable gat-ing mechanism that enables gradient-based training. This formulation adapts the screening hyperplane to the investor’s return goal, effectively coupling asset selection with allocation. Empirical results on sector-diversified U.S. equities demonstrate that the integrated model delivers improved risk–return trade-offs over the two-stage baseline, with pronounced gains in volatile market regimes, representing the first differentiable integration of SVM and MVO.
- 6 July 2025
- Published by Springer Nature in Computational Management Science
- Vol. 22 (2) , 1-35
A novel regime-switching commodity pricing model with stochastic convenience yield
- 6 July 2025
- Published by Springer Nature in Computational Management Science
- Vol. 22 (2) , 1-35
A novel regime-switching commodity pricing model with stochastic convenience yield
- 4 March 2025
- Published by Springer Nature in Journal of Global Optimization
- Vol. 91 (4) , 923-951
Stochastic red-blue set covering: a decomposition approach
- 4 March 2025
- Published by Springer Nature in Journal of Global Optimization
- Vol. 91 (4) , 923-951
Stochastic red-blue set covering: a decomposition approach
- Open Access
Contextual Scenario Generation for Two-Stage Stochastic Programming
Two-stage stochastic programs (2SPs) are important tools for making decisions under uncertainty. Decision-makers use contextual information to generate a set of scenarios to represent the true conditional distribution. However, the number of scenarios required is a barrier to implementing 2SPs, motivating the problem of generating a small set of surrogate scenarios that yield high-quality decisions when they represent uncertainty. Current scenario generation approaches do not leverage contextual information or do not address computational concerns. In response, we propose contextual scenario generation (CSG) to learn a mapping between the context and a set of surrogate scenarios of user-specified size. First, we propose a distributional approach that learns the mapping by minimizing a distributional distance between the predicted surrogate scenarios and the true contextual distribution. Second, we propose a task-based approach that aims to produce surrogate scenarios that yield high-quality decisions. The task-based approach uses neural architectures to approximate the downstream objective and leverages the approximation to search for the mapping. The proposed approaches apply to various problem structures and loosely only require efficient solving of the associated subproblems and 2SPs defined on the reduced scenario sets. Numerical experiments demonstrating the effectiveness of the proposed methods are presented.
Open AccessContextual Scenario Generation for Two-Stage Stochastic Programming
Two-stage stochastic programs (2SPs) are important tools for making decisions under uncertainty. Decision-makers use contextual information to generate a set of scenarios to represent the true conditional distribution. However, the number of scenarios required is a barrier to implementing 2SPs, motivating the problem of generating a small set of surrogate scenarios that yield high-quality decisions when they represent uncertainty. Current scenario generation approaches do not leverage contextual information or do not address computational concerns. In response, we propose contextual scenario generation (CSG) to learn a mapping between the context and a set of surrogate scenarios of user-specified size. First, we propose a distributional approach that learns the mapping by minimizing a distributional distance between the predicted surrogate scenarios and the true contextual distribution. Second, we propose a task-based approach that aims to produce surrogate scenarios that yield high-quality decisions. The task-based approach uses neural architectures to approximate the downstream objective and leverages the approximation to search for the mapping. The proposed approaches apply to various problem structures and loosely only require efficient solving of the associated subproblems and 2SPs defined on the reduced scenario sets. Numerical experiments demonstrating the effectiveness of the proposed methods are presented.