Ji Huang
  • Home
  • Working Papers
  • Publications
  • Teaching
Deep Learning-Based Probabilistic Approach
     
all implementing codes below available upon request

A Probabilistic Solution to High-Dimensional Continuous-Time Macro and Finance Models, Jun 2025, slides, 
   keywords: Forward-Backward SDE, deep learning, heterogeneous-agent, flow of conditional distribution

Breaking the Curse of Dimensionality in Heterogeneous-Agent Models: A Deep Learning-Based Probabilistic Approach, May 2024,
   keywords: two-asset portfolio choice, nonlinear market clearing condition, search and matching frictions,
​                      over-the-counter markets
    recent presentation: SED 2024 Barcelona, Sep24 NUS, Dec24 Peking U

​
Applications of Deep Learning-Based Probabilistic Approach to "Combinatorial" Problems in Economics (w/ Jinghai Yu), Aug 2024,
   keywords: combinatorial discrete choice, dynamic network formation
   recent presentation: EEA-ESEM 2024 Rotterdam, ASSA 2025 San Francisco


Applications of Deep Learning-Based Probabilistic Approach to Economic Models with High-Dimensional Controls (w/ Hongxiao Chen), Apr 2025
   keywords: debt maturity management, preferred habitat models, term structure of interest rate 

Macro and Finance
Bond Finance, Bank Finance, and Bank Regulation, Apr 2020

The Economics of Consumer Credit on an E-Commerce Platform, Aug 2022
   keywords: return to scope, boundary of a platform, BNPL, FinTech 

Endogenous Global Risk (w/ Hao Hu), Jan 2022
   keywords: global financial cycle, exorbitant privilege, valuation effect, global saving glut, financial contagion
  
Less E-Commerce Platform Competition Benefits Financial Inclusion (w/ Zehao Zheng), Jun 2022

The Implicit Guarantee Channel: How Policy Uncertainty Affects Shadow Banks (w/ Zongbo Huang and Xiang Shao), Nov 2023

Dormant Projects
Systemic Run on Shadow Banks: the Minsky Moment, Aug 2014

​
​A Tale of Two Items Sent Off Balance Sheet: Securitization and Loan Commitment, Jan 2017​
Powered by Create your own unique website with customizable templates.