Browse our site
About
People
Research Areas
Projects
Publications
Seminars
Future
Past
News
You are here:
Home
Seminars
Past
Interest Rate Model Calibration using Gaussia...
Interest Rate Model Calibration using Gaussian Processes for Machine Learning
Main information
By:
João Beleza Sousa (
Instituto Superior de Engenharia de Lisboa
)
Date:
Wednesday, 18th of February 2009, 14h00
Location:
FCT/UNL, Seminar Room (Ed. II)
Abstract
With Kernel Machines playing a central role in Machine Learning, models based on Gaussian Processes have become increasely popular for problems of regression and classification.
A problem of learning with Gaussian Processes is the specification of mean and covariance functions to use. This requires detailed prior information on the data which in many problems is not known.
In the field of Financial Mathematical Modeling some Gaussian processes are extensively used because they proved to be good models for some prices sequences, and also because of their analytical simplicity. Their mean and covariance functions are analytically computable.
A problem of using such models is the calibration problem. What parameters to use given historical and present prices.
In this talk we connect the worlds of Machine Learning and Financial Mathematical Modeling.
We calibrate (learn the parameters) of a Gaussian interest rate model (the Vasicek model) with Gaussian Processes for Machine Learning using the mean and covariance functions computed analytically.
Short-bio
João Beleza Sousa is Adjunct Professor at Instituto Superior de Engenharia de Lisboa (ISEL), where he teaches Digital Signal Processing.
His past experience includes: teaching Stochastic Processes at ISEL, web systems design and development for L. J. Carregosa Stock Broker, and research in the area of automatic speech recognition at Instituto de Engenharia de Sistemas e Computadores (INESC).
He holds master degree in electrical and computer engineering from Instituto Superior Técnico. Currently he is researching in the area of Machine Learning Methods for Financial Mathematical Modeling in collaboration with Prof. Manuel Esquível from Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa.
Departamento de Informática, FCT/UNL
Quinta da Torre 2829-516 CAPARICA - Portugal
Tel. (+351) 21 294 8536 FAX (+351) 21 294 8541
