TalksWhen the Model is No Longer Trustworthy - A Distribution - Free Optimization Approach
99
reads
In the past decades, the financial industry has experienced several catastrophic losses such as a hundred million loss reported by Bank of Tokyo/Mitsubishi in 1997, and the unprecedented financial crisis in 2008 that rocked the global economy. These losses have been largely attributed to deficiency in risk management and inadequacy of the underlying probabilistic model used to capture the stochastic nature of the markets. These recurring crises shed light on the devastating impact of over-reliance on a single probabilistic model in financial decision making. In this talk, I will present a modern optimization approach, Distributionally Robust Optimization (DRO) that allows decisions to be optimized by further taking into account the ambiguity in the choice of a probabilistic model. In particular, DRO provides decision makers the flexibility to specify the underlying model using only partial statistical information. I will review the theoretical property and the computational advantage of DRO, and illustrate the benefit of using DRO via its application in portfolio optimization and derivative pricing. Latest developments of DRO will also be addressed.
reads
2011-09-09
11:00:00 - 12:00:00
400-1 , Mathematics Research Center Building (ori. New Math. Bldg.)
In the past decades, the financial industry has experienced several catastrophic losses such as a hundred million loss reported by Bank of Tokyo/Mitsubishi in 1997, and the unprecedented financial crisis in 2008 that rocked the global economy. These losses have been largely attributed to deficiency in risk management and inadequacy of the underlying probabilistic model used to capture the stochastic nature of the markets. These recurring crises shed light on the devastating impact of over-reliance on a single probabilistic model in financial decision making. In this talk, I will present a modern optimization approach, Distributionally Robust Optimization (DRO) that allows decisions to be optimized by further taking into account the ambiguity in the choice of a probabilistic model. In particular, DRO provides decision makers the flexibility to specify the underlying model using only partial statistical information. I will review the theoretical property and the computational advantage of DRO, and illustrate the benefit of using DRO via its application in portfolio optimization and derivative pricing. Latest developments of DRO will also be addressed.