Tensor Processing Units for Financial Monte Carlo

被引:0
|
作者
Belletti, Francois [1 ]
King, Davis [1 ]
Yang, Kun [1 ]
Nelet, Roland [1 ]
Shafi, Yusef [1 ]
Shen, Yi-Fan [1 ]
Anderson, John [1 ]
机构
[1] Google Res, Mountain View, CA 94043 USA
关键词
DERIVATIVES; SIMULATION;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Monte Carlo methods are critical to many routines in quantitative finance such as derivatives pricing, hedging and risk metrics. Unfortunately, Monte Carlo methods are very computationally expensive when it comes to running simulations in high-dimensional state spaces where they are still a method of choice in the financial industry. Recently, Tensor Processing Units (TPUs) have provided considerable speedups and decreased the cost of running Stochastic Gradient Descent (SGD) in Deep Learning. After highlighting computational similarities between training neural networks with SGD and simulating stochastic processes, we ask in the present paper whether TPUs are accurate, fast and simple enough to use for financial Monte Carlo. Through a theoretical reminder of the key properties of such methods and thorough empirical experiments we examine the fitness of TPUs for option pricing, hedging and risk metrics computation. In particular we demonstrate that, in spite of the use of mixed precision, TPUs still provide accurate estimators which are fast to compute when compared to GPUs. We also show that the Tensorflow programming model for TPUs is elegant, expressive and simplifies automated differentiation.
引用
收藏
页码:12 / 23
页数:12
相关论文
共 50 条
  • [21] Monte Carlo grid for financial risk management
    Tezuka, S
    Murata, H
    Tanaka, S
    Yumae, S
    FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE, 2005, 21 (05): : 811 - 821
  • [22] Monte Carlo methods for pricing financial options
    Bolia, N
    Juneja, S
    SADHANA-ACADEMY PROCEEDINGS IN ENGINEERING SCIENCES, 2005, 30 (2-3): : 347 - 385
  • [23] Monte Carlo methods for pricing financial options
    N. Bolia
    S. Juneja
    Sadhana, 2005, 30 : 347 - 385
  • [24] Monte Carlo methids for signal processing
    Doucet, A
    Wang, XD
    IEEE SIGNAL PROCESSING MAGAZINE, 2005, 22 (06) : 152 - 170
  • [25] Tensor Monte Carlo: Particle Methods for the GPU era
    Aitchison, Laurence
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 32 (NIPS 2019), 2019, 32
  • [26] Toward large-scale Hybrid Monte Carlo simulations of the Hubbard model on graphics processing units
    Wendt, Kyle A.
    Drut, Joaquin E.
    Lahde, Timo A.
    COMPUTER PHYSICS COMMUNICATIONS, 2011, 182 (08) : 1651 - 1656
  • [27] Parallel computing with graphics processing units for high-speed Monte Carlo simulation of photon migration
    Alerstam, Erik
    Svensson, Tomas
    Andersson-Engels, Stefan
    JOURNAL OF BIOMEDICAL OPTICS, 2008, 13 (06)
  • [28] Aging at the spin-glass/ferromagnet transition: Monte Carlo simulations using graphics processing units
    Manssen, Markus
    Hartmann, Alexander K.
    PHYSICAL REVIEW B, 2015, 91 (17)
  • [29] MARKOV CHAIN MONTE CARLO METHODS IN FINANCIAL ECONOMETRICS
    Verhofen, Michael
    FINANCIAL MARKETS AND PORTFOLIO MANAGEMENT, 2005, 19 (04) : 397 - 405
  • [30] Reconfigurable acceleration for Monte Carlo based financial simulation
    Zhang, GL
    Leong, PHW
    Ho, CH
    Tsoi, KH
    Cheung, CCC
    Lee, DU
    Cheung, RCC
    Luk, W
    FPT 05: 2005 IEEE INTERNATIONAL CONFERENCE ON FIELD PROGRAMMABLE TECHNOLOGY, PROCEEDINGS, 2005, : 215 - 222