Search this site
Embedded Files
SKKU MLFE lab
  • Home
  • Research (PG-DPO)
    • Why BPTT ≈ Costate?
  • People
    • Principal Investigator
      • Talks
      • Teaching
  • Papers
    • Papers in Progress
  • Projects
SKKU MLFE lab
  • Home
  • Research (PG-DPO)
    • Why BPTT ≈ Costate?
  • People
    • Principal Investigator
      • Talks
      • Teaching
  • Papers
    • Papers in Progress
  • Projects
  • More
    • Home
    • Research (PG-DPO)
      • Why BPTT ≈ Costate?
    • People
      • Principal Investigator
        • Talks
        • Teaching
    • Papers
      • Papers in Progress
    • Projects

Back to Papers

Preprints

  • Breaking the Dimensional Barrier: A Pontryagin-Guided Direct Policy Optimization for Continuous-Time Multi-Asset Portfolio, Jeonggyu Huh*, Jaegi Jeon, Hyeng-Keun Koo, Byung-Hwa Lim* [Paper] [Code] [Slides]  -> o
    + Supplementary proofs: Detailed  Proof of Theorem 3 — BPTT–BSDE Equivalence [PDF] · Detailed  Proof of Theorem 4 — Policy-Gap Bounds [PDF] 

  • Breaking the Dimensional Barrier: Dynamic Portfolio Choice with Parameter Uncertainty via Pontryagin Projection, Jeonggyu Huh*, Hyeng-Keun Koo [Paper] [Slide] 

  • From Value Bounds to Policy-Distance and Active-Face Certificates: Same-Grid Duality for Constrained Dynamic Portfolios [Link]

  • A Computational Framework for Decision-Aligned Conditional Betas in Cost-Aware Portfolio Optimization, Dongwan Shin*, Hojin Ko, Jeonggyu Huh*

Papers in Progress

>> Stochastic Control 

  • Structure-Preserving Policy Iteration for Dynamic Portfolio Choice: Whole-Space Convergence, Localization, and Neural Certification, with Yeoneung Kim, Seungwon Jeong

  • Finite Horizon and Optimal Portfolio Choice with Stochastic Income: A Reinforcement Learning Approach, with Seyoung Park, Hojin Ko, Alain Bensoussan-> o

  • Pontryagin-Guided Deep Hedging, with Seungho Na, Hojin Ko 

  • Recovering the Kinks in Delay Control: A Structure-Aware Optimal Control Solver with Pontryagin Projection, with Seungwon Jeong, Jihun Kim

  • Scalable Distributionally Robust Portfolio Choice via Pontryagin-Guided Direct Policy Optimization, with Ho-Jun Lee


>> DeepONet

  • Real-Time Pricing of Equity-Linked Securities Using Deep Operator Networks, with Yoonyoung Byun, Jae Wook Song, Juhwan Kim


>> Asset Pricing

  • Discounted Alpha: A Machine Learning Framework for Equity Valuation, with Dongwan Shin, Thummim Cho

Future directions 

  • Beyond its current scope, PG-DPO admits natural extensions to partial equilibrium settings—including robust optimization, rough volatility, regime, taxation, Epstein-Zin utility, smooth ambiguity and belief-state dynamics—as well as general equilibrium frameworks such as mean-field control and multi-agent games.

Google Sites
Report abuse
Google Sites
Report abuse