PARALLELISM STRATEGIES FOR NEUROPHYSIOLOGICAL DELAYED TRANSFER ENTROPY DATA PROCESSING: TOWARDS CAUSAL INFERENCE IN BIG DATA

JONAS  ROSSI DOURADO; MICHEL  BESSANI; DANIEL RODRIGUES DE LIMA; JOSÉ ROBERTO  B. DE A. MONTEIRO; RAFAEL  RODRIGUES MENDES RIBEIRO; CARLOS  DIAS MACIEL

doi:10.20906/CBA2022/429

JONAS ROSSI DOURADO Universidade de São Paulo
MICHEL BESSANI UNIVERSIDADE DE SÃO PAULO
DANIEL RODRIGUES DE LIMA UNIVERSIDADE DE SÃO PAULO
JOSÉ ROBERTO B. DE A. MONTEIRO UNIVERSIDADE DE SÃO PAULO
RAFAEL RODRIGUES MENDES RIBEIRO UNIVERSIDADE DE SÃO PAULO
CARLOS DIAS MACIEL UNIVERSIDADE DE SÃO PAULO

DOI: https://doi.org/10.20906/CBA2022/429

Keywords: Delayed Transfer Entropy, Neurophysiological Data, Causality, Parallelism Strategies, Big Data Analysis, Task Parallelism, Data Parallelism, Data Analysis Performance

Abstract

Nowadays, the amount of data being generated and collected has been rising with the popularization of technologies such as Internet of Things, social media, and smartphone. The increasing amount of data led the creation of the term big data. One class of Big Data hidden information is causality. Among the tools to infer causal relationships there is Delayed Transfer Entropy (DTE); however, it has a high demanding processing power. Many approaches were proposed to overcome DTE performance issues such as GPU and FPGA implementations. Our approach is to compare different parallel strategies to calculate DTE from neurophysiological time series using a heterogeneous Beowulf cluster aiming to increase DTE performance.