Big Data Analytics. Tools and Techniques - PDF

Please download to get full document.

View again

of 14
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Bills

Published:

Views: 7 | Pages: 14

Extension: PDF | Download: 0

Share
Related documents
Description
Big Data Analytics Basic concepts of analyzing very large amounts of data Dr. Ing. Morris Riedel Adjunct Associated Professor School of Engineering and Natural Sciences, University of Iceland Research
Transcript
Big Data Analytics Basic concepts of analyzing very large amounts of data Dr. Ing. Morris Riedel Adjunct Associated Professor School of Engineering and Natural Sciences, University of Iceland Research Group Leader, Juelich Supercomputing Centre, Germany LECTURE BDA5 Tools and Techniques (v2) Online Material Outline Lecture BDA5 Tools and Techniques 2/ 14 Outline Tools Overview and Selected Use Cases RMPI R interface to MPI Methods Statistical Techniques Data Mining Machine Learning Evolutionary Optimization Lecture BDA5 Tools and Techniques 3/ 14 RMPI Lecture BDA5 Tools and Techniques 4/ 14 RMPI Overview The tool R is a free software environment for statistical computing and graphics RMPI is an R package and an interface (wrapper) to the Message Passing Interface (MPI) MPI is a standardized and portable parallel programming model implemented via libraries RMPI ports low level MPI functions into R so that users do not have to know C or Fortran MPI libraries supported: LAM-MPI, MPICH(2), and OpenMPI Write R programs using certain RMPI functions: Startup and Shutdow: e.g. mpi.spawn.rslaves([nslaves=#]) Cluster Information: e.g. mpi.comm.rank() Use sending/receiving data/functions to send an R object like a number, string, or a list between different slaves or to all slaves (e.g. mpi.bcast.robj2slave(object)) [2] RMPI Web Page Enables inter process communication Lecture BDA5 Tools and Techniques 5/ 14 RMPI Selected Use Cases Precision Medicine [1] Bottomly et al., 2013 Simulations and calculation of P-values and False Discovery Rate (FDR) Parallel computed on Beowulf-style cluster using RMPI R Use of parallel random number generation using L Ecuyer s method (R rlecuyer package) Plotting and summaries for the simulations with ggplot2 on R Lecture BDA5 Tools and Techniques 6/ 14 Methods Lecture BDA5 Tools and Techniques 7/ 14 Statistical Techniques The main task of statistical techniques is to test the likelihood of a given hypothesis Limits Impossible to identify new relationships in data Discovery of new relationships is a very important chance to realize additional unexpected benefits from the large amounts of data Example: data available from High Throughput Experimentation (HTE) [3] Ohrenberg et al., 2005 Lecture BDA5 Tools and Techniques 8/ 14 Data Mining Techniques Data mining stands for algorithms and techniques which transfer data into information Provides methods that extract rules, patterns, or other structural information from data Identify in the data new relationships that were previously undiscovered Typical data-mining techniques Clustering, separation methods, decision trees, association rules, etc. Limits Data mining does not assess the validity of extracted rules Making data-mining results reliable, it is essential to apply different approaches and check the resulting rules for mutual compatibility [3] Ohrenberg et al., 2005 Lecture BDA5 Tools and Techniques 9/ 14 Machine Learning Techniques Machine learning techniques are able to learn from data and train a system (model) Wide variety of machine learning algorithms exist Artificial Neural Networks, Support Vector Machines, etc. Artificial Neural Networks (ANN) A kind of extremely powerful regression model (rough perspective) Input-output relationships can be generated by ANN as black box model Black box model: function cannot be interpreted physically the parameters in the ANN have generally no explicit physical meaning ANN need less data then other black-box methods to generate an input output relationship of comparible quality (given a wide range of conditions) [3] Ohrenberg et al., 2005 Lecture BDA5 Tools and Techniques 10 / 14 Evolutionary Optimization Evolutionary optimization techniques are characterized by a stochastic approach Advantages Compared to statistical experiment design, evolutionary algorithms are very flexible and adaptable Limits In most cases the optimization speed is rather slow [3] Ohrenberg et al., 2005 Lecture BDA5 Tools and Techniques 11 / 14 Lecture Bibliography Lecture BDA5 Tools and Techniques 12 / 14 Lecture Bibliography [1] Bottomly et al.: Comparison of methods to identify aberrant expression patterns in individual patients: augmenting our toolkit for precision medicine. Genome Medicine :103., Online: [2] RMPI Web Page, Online: [3] Ohrenberg, Arne, et al. Application of Data Mining and Evolutionary Optimization in Catalyst Discovery and High-Throughput Experimentation Techniques, Strategies, and Software. QSAR & Combinatorial Science 24.1 (2005): Lecture BDA5 Tools and Techniques 13 / 14 Lecture BDA5 Tools and Techniques 14 / 14
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x