Home > Events > 2013 Seminars & Colloquia > Subhadeep (Deep) Mukhopadhyay - Texas A & M University

Subhadeep (Deep) Mukhopadhyay - Texas A & M University

Main Content

"United Statistical Algorithms for Small & Big Data"
17 January 2013 from 4:00 PM to 5:00 PM
201 Thomas Bldg.
Add event to calendar

Some of the important characteristics of real-world modern data sets: (i) data structure:

univariate, bivariate, time series, matrix valued, text, image, network etc.; (ii)  data type: continuous,  discrete, categorical, censored, missing etc.; (iii)  data quality:  often noisy and hidden nonlinearity; (iv) data size: small p large n, small p small n, large p small n, and large p large n.

In this talk, I will introduce a new general framework called United Statistical Algorithm to tackle the various aforementioned aspects of recent data in a more systematic and coherent way. The plan is to use series of examples to present the big picture. The proposed LP-algorithm unifies many cultures of statistical science and statistical learning methods. Modern Nonparametric Statistics incorporates traditional methods and made it relevant to modern Big data, categorical data analysis, network analysis, regression, classification, time series etc.  Our primary goal is to design interpretable and scalable algorithms that integrate confirmatory and exploratory inference. I will emphasize the fundamental role played by our specially designed (data-specific) mid-distribution based Legendre like orthonormal Polynomials for parsimonious representation of complex data.

This is joint work with Professor Manny Parzen.

Keywords and phrases: Copula, Custom made Orthogonal basis function, Mid-distribution function, Comparison density, Contingency table, MaxEnt, L2-estimation, Sufficient statistics, Big data, Quantile function, Sparse modeling,  Information divergence, LP moments and Comoments, LPINFOR, Dimension reduction, Singular value decomposition.

Alternative Title: An Introduction to Data Modeling using “LP-STAT”.

Filed under: ,