Dear All,
I cordially invite you to the Data Science and Statistics Seminar in Math Dept (Ayres 111), today Thursday 11/9 @4:30PM.
Our speaker will be Dr. Tomojit Ghosh from UT Chattanooga who will introduce an autoencoder based on a supervised approach for data viz. The auto encoder relies on a pertinent sparse optimization, and will be applied on biological data sets.
For more information, please see below.
Thank you and I am looking forward to seeing you today at 4:30PM.
Vasileios Maroulas
Speaker: Tomojit Ghosh
Affiliation: University of Tennessee Chattanooga
Title: Some Convex and Nonconvex Models to Analyze Biological Data Sets.
Abstract: Classification, visualization, and feature selection are three essential tasks of machine learning. This talk presents convex and non-convex models suitable for these three tasks, especially for high-dimensional biological data sets. We introduce Centroid-Encoder (CE), an autoencoder-based supervised data visualization tool. We further developed a sparse optimization problem for the non-linear mapping of the centroid-encoder, Sparse Centroid-Encoder (SCE), to determine the set of discriminative features between two or more classes. CE and SCE are models based on neural network architectures and require the solution of non-convex optimization problems. Motivated by the CE algorithm, we present a linear formulation of the Centroid-Encoder with orthogonality constraints called Linear Centroid Encoder (LCE). This formulation is similar to PCA (Principal Component Analysis), except the class labels are used to formulate the objective, resulting in supervised PCA. Finally, we present a linear formulation of the Sparse Centroid-Encoder called Sparse Linear Centroid Encoder (SLCE), which employs a two-step training process where each step is convex. The author will show the application of these algorithms on biological data sets.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
To leave the list go here:
https://listserv.utk.edu/cgi-bin/wa?SUBED1=MATHTALK&A=1
|
