Third Year Candidacy Requirement: "Prediction theory for stationary Time Series"

Abstract: Nonlinear prediction of time series can offer potential accuracy gains over linear methods when the process is non-linear. As there are numerous examples of non-linearity in time-series data (e.g., finance, macroeconomics, image, and speech processing), there seems to be merit in developing a general theory and methodology. In this presentation I am going to give a brief overview of time series analysis, existing non-linear prediction techniques, a linearity testing methodology and a short overview of our work on quadratic prediction and polyspectral mean.

Host: Soumendra Lahiri