Thu.1 09:00–10:15 | H 3007 | BIG
.

Big Data and Machine Learning – Contributed Session 1/3

Chair: Guozhi Dong
9:00

Scindhiya Laxmi

joint work with S. K. Gupta

Human Activity Recognition using Intuitionistic Fuzzy Proximal Support Vector Machines

Human activity recognition is an active area of research in Computer Vision. One of the difficulties of an activity recognition system is the presence of noises in the training data. We addressed this problem by assigning membership and hesitation degrees to the training instances according to its contribution to the classification problem. In this way, the impact of noises and outliers has been reduced to some extent. The methodology intuitionistic fuzzy proximal support vector machine for human activity is proposed to speed up the training phase and to increase the classification accuracy.

9:25

Robert Ravier

joint work with Vahid Tarokh

Online Optimization for Time Series of Parametrizable Objective Functions

The time-varying component of many objective functions of practical interest in online optimization is fully contained in a finite number of parameters. This allows us to use methods from standard time series analysis in order to predictively optimize. We investigate the theoretical and computational effects of using these methods. In particular, we focus on both the potential improvements of regret upper bounds when utilizing these techniques as well as algorithms for simultaneously modeling and optimizing.

9:50

Guozhi Dong

joint work with Michael Hintermüller, Kostas Papafitsoros

A data-driven model-based method for quantitative MRI

Recently, an integrated physics-based model has been introduced for quantitative magnetic resonance imaging. It incorporates the physical model by Bloch equations into the data acquisition, and leads to high accuracy estimation of the tissue parameters. However, in general, the physical model might be not explicitly known. In this talk, we introduce a novel data-driven model-based method which combines the advances of integrated physics-based method with deep neural networks.