High Performance Computing in Neuroimaging using BROCCOLI. Anders Eklund, PhD Virginia Tech Carilion Research Institute

Size: px
Start display at page:

Download "High Performance Computing in Neuroimaging using BROCCOLI. Anders Eklund, PhD Virginia Tech Carilion Research Institute"

Transcription

1 High Performance Computing in Neuroimaging using BROCCOLI Anders Eklund, PhD Virginia Tech Carilion Research Institute

2 Outline What is BROCCOLI? What is high performance computing and GPUs? Why are GPUs needed in neuroimaging? GPUs in medical imaging, neuroscience, statistics, machine learning What can BROCCOLI do? Demos Future

3 What is BROCCOLI? BROCCOLI is a new software package for fmri analysis The main advantage of BROCCOLI is that it is much faster compared to other fmri softwares BROCCOLI performs all calculations in parallel and can, for example, run on CPUs, Nvidia GPUs and AMD GPUs

4 Is a new software package necessary? The importance of processing package development has been underrated. SPM, FSL, AFNI all fundamentally push the field as they are distributed. The field progress is directly related to how dynamic and cutting edge these packages are since we all rely on them. Peter Bandettini, NSF: Neuroimaging Final Workshop, May 23, 2014

5 What is high performance computing? Aggregate computing power to obtain much higher performance, than possible with a regular computer Used to solve large problems in science & engineering PC clusters (many computers) GPUs (graphics cards)

6 Graphics card What is a GPU? Graphics processing unit The part of a graphics card that performs calculations Calculations are done in parallel, for example to render a large number of pixels at interactive frame rates GPU

7 A poor man s supercomputer (~3000 USD)

8 Giga floating point operations per second (GFLOPS) CPU vs GPU Theoretical Performance 8 From the Nvidia CUDA Programming Guide

9 Serial vs parallel computing Adding two vectors, N elements Serial computing, one thread performs all operations, one at a time (C) Parallel computing, N threads, run in parallel (CUDA) for (int i = 0; i < N; i++) { } C[i] = A[i] + B[i]; int i = blockidx.x * blockdim.x + threadidx.x; if (i < N) C[i] = A[i] + B[i];

10 Why are GPUs needed in neuroimaging? Higher temporal and spatial resolution More subjects Demanding image processing algorithms Demanding statistical algorithms Demanding machine learning algorithms

11 Why are GPUs needed in neuroimaging? Higher temporal and spatial resolution Standard resolution dataset (4 x 4 x 4 mm, 0.5 Hz), 64 x 64 x 30 x 430 High resolution dataset (2 x 2 x 2 mm, 1.4 Hz), 104 x 90 x 72 x times more data!

12 Why are GPUs needed in neuroimaging? More subjects Data sharing initiatives enable large scale analyses OpenfMRI.org, 420 subjects 1000 Functional Connectomes Project, 1200 subjects Human Connectome Project, goal 1200 subjects

13 Why are GPUs needed in neuroimaging? Demanding image processing algorithms Image registration, to align two images or volumes Image segmentation, to extract a certain part of a dataset Image denoising, to suppress noise and improve image quality

14 Why are GPUs needed in neuroimaging? Demanding statistical algorithms Non-parametric statistics (e.g. permutation tests, bootstrap, jackknife) Cross validation Markov Chain Monte Carlo (MCMC) simulation Dynamic causal modeling

15 Why are GPUs needed in neuroimaging? Demanding machine learning algorithms Support vector machines Search light, a classifier in each voxel Independent component analysis Deep learning Bayesian methods

16 Why are GPUs needed in neuroimaging? Human connectome project; 1200 subjects, 3 GB of resting state data per run, 4 runs Apply a permutation test (10,000 permutations) to each subject and each run Equivalent to analyze 144 petabytes of data (1200 * 3 GB * 4 * = GB)

17 How are GPUs being used in fields related to neuroimaging?

18 GPUs in Medical Imaging, Modalities Eklund et al., Medical Image Processing on the GPU Past, present and Future, Medical Image Analysis, 2013

19 GPUs in Medical Imaging, Algorithms Eklund et al., Medical Image Processing on the GPU Past, present and Future, Medical Image Analysis, 2013

20 GPUs in Neuroscience Neural field models, nonlinear integral differential equations Input Stimulus Simulating stochastic networks Estimating network probability distributions Output Brain Activity Baladron et al., Three Applications of GPU Computing in Neuroscience, Computing in Science & Engineering, 2012

21 GPUs in Statistics Non-linear regression Resampling based methods (bootstrap, permutation, jackknife, cross-validation) Markov Chain Monte Carlo (MCMC) simulation Random number generation directly on the GPU Suchard et al., Understanding GPU programming for statistical computation: Studies in massively parallel massive mixtures, Journal of Computational and Graphical Statistics, 2010

22 The Google Brain Project Deep learning using 1.7 billion parameters 10 million YouTube video thumbnails, 200 x 200 pixels No labels, unsupervised learning 16,000 CPU cores (1000 machines), 3 days of processing Recognizes human faces and cat faces Le, Q., Building high-level features using large scale unsupervised learning, ICASSP, 2013

23 GPUs in Machine Learning Used 3 GPU machines instead of ,432 GPU cores $30,000 instead of $5,000,000 4 kwatts instead of 600 kwatts Same processing time Coates et al., Deep learning with COTS HPC Systems, JMLR, 2013

24 GPUs vs Supercomputers GPUs are much cheaper GPUs are much more power efficient GPUs have a much smaller physical size GPUs enable local data storage Do not have to share the resources with anyone else

25 Why are GPUs uncommon in neuroimaging?

26 Software packages for fmri analysis No software package with GPU support for fmri analysis Existing softwares are not prepared for the medical data explosion and demanding algorithms BROCCOLI is intended to fill this gap

27 How can a GPU be programmed?

28 CUDA CUDA = Compute Unified Device Architecture Released in 2007, based on the C programming language Several highly optimized libraries (e.g. CUFFT, CURAND, CUBLAS, CUSPARSE) Only works for Nvidia GPUs 28

29 OpenCL OpenCL = Open Computing Language Works for many types of hardware CPUs, GPUs, field programmable gate arrays (FPGAs), digital signal processors (DSPs), Intel Xeon Phi, Not as many libraries as CUDA (yet) 29

30 BROCCOLI uses OpenCL Can use both Nvidia and AMD GPUs Can also run the same code on a CPU Not necessary to buy a GPU to try BROCCOLI!

31 BROCCOLI library Ongoing Future

32 BROCCOLI bash functions GetOpenCLInfo, lists all OpenCL platforms and devices RegisterTwoVolumes, linear and non-linear registration FirstLevelAnalysis, including permutation testing and MCMC RandomiseGroupLevel, permutation testing MotionCorrection

33 Demo 1, OpenCL platforms and devices

34 Image registration in BROCCOLI Most image registration algorithms are based on optimizing an intensity based similarity measure (correlation, mutual information, ) BROCCOLI instead relies on the concept of local phase from quadrature filters

35 Quadrature filters Spatial domain Line detector (even) Real part Edge detector (odd) Imaginary part Knutsson, H., What s so good about quadrature filters? ICIP, 2003

36 Quadrature filters Frequency domain Knutsson, H., What s so good about quadrature filters? ICIP, 2003

37 Local phase

38 Minimizing the optical flow equation The linear registration algorithm in BROCCOLI is based on minimizing the optical flow equation The equation cannot be solved, since there is one equation and three motion parameters per voxel

39 Minimizing the optical flow equation Use a parametric model of the motion field

40 Minimizing the optical flow equation Put the parametric model into the error expression Calculate the derivative, with respect to p

41 Minimizing the optical flow equation Set the derivative to 0, to find the minimum Solve the equation system to get the best parameters, no optimization algorithm required

42 Phase based optical flow Replace intensity difference with phase difference Replace image gradient with phase gradient Add a certainty for each voxel (e.g. the magnitude of the quadrature filter response)

43 Linear registration algorithm Apply 3 quadrature filters (along x, y and z) Calculate phase differences, phase gradients and certainties in each voxel Setup a linear equation system for the entire image Solve the equation system to get the best transformation parameters Apply transformation parameters using interpolation Iterate

44 Non-linear registration The non-linear registration algorithm used in BROCCOLI is called the Morphon The Morphon also uses the concept of local phase Solve one equation system in each voxel, instead of one equation system for the entire image Knutsson & Andersson, Morphons: Segmentation Using Elastic Canvas and Paint on Priors, ICIP, 2005

45 The Morphon L2 norm minimization in each voxel Structure tensor Filter direction Certainty Phase difference Displacement vector

46 The local structure tensor T Describes the strength and the orientation of the signal in an image or a volume Analogous to the tensor in diffusion tensor imaging Apply filters along different directions (at least 6 in 3D) and combine the filter responses to a tensor Knutsson, H., Representing local structure using tensors, SCIA, 1989

47 The Morphon Calculate the derivative with respect to d, set it to 0, solve the equation system, no optimization algorithm required

48 Scale pyramid Find approximate solution at a low resolution Improve the solution by moving to finer and finer scales Downsampling a factor 8, 4, 2, 1

49 RegisterTwoVolumes - Bash RegisterTwoVolumes inputvolume.nii referencevolume.nii [options] Some of the options are -platform -device -lowestscale OpenCL platform to use OpenCL device to use Highest downsampling factor

50 Demo 2, linear and non-linear registration in BROCCOLI

51 Demo 2, Summary Can easily change the used platform Linear and non-linear registration to a 2 mm MNI template in 1 second Linear and non-linear registration to a 1 mm MNI template in 5 seconds The registration is rather quick even with a CPU

52 Demo 2, Summary Non-linear registration of 200 T1 volumes to a brain template (182 x 218 x 182) FSL FSL, 4 terminals AFNI AFNI OpenMP BROCCOLI Intel CPU BROCCOLI Nvidia GPU BROCCOLI AMD GPU 116 hours 29 hours 110 hours 32 hours 3.5 hours 15 minutes 20 minutes

53 Average similarity metrics from 200 registrations Correlation Mutual information Sum of squared differences

54 First level analysis in BROCCOLI Registration of T1 volume to a brain template (using linear and non-linear registration) Registration of one fmri volume to the T1 volume Motion correction Smoothing Statistical analysis

55 FirstLevelAnalysis - Bash FirstLevelAnalysis fmri.nii T1.nii braintemplate.nii regressors.txt contrasts.txt [options] Some of the options are -platform -device -permute -bayesian OpenCL platform to use OpenCL device to use Apply permutation test Run MCMC simulation

56 Demo 3, first level analysis in BROCCOLI

57 Demo 3, Summary First level analysis in ~10 seconds, for a 2 mm MNI template First level analysis in ~25 seconds, for a 1 mm MNI template

58 Demo 3, Summary Intermediate results take up a lot of disk space Original dataset 500 MB, motion corrected 500 MB, smoothed 500 MB, Unnecessary to store all intermediate results if BROCCOLI is used, recalculate them if necessary

59 How should the statistical maps be thresholded? How should we correct for multiple testing?

60 Parametric vs non-parametric statistics Most fmri software packages rely on parametric statistical algorithms to calculate p-values The parametric algorithms are based on a number of assumptions about the data These assumptions become more critical when we correct for multiple testing

61 Eklund et al., Does parametric fmri analysis with SPM yield valid results? An empirical study of 1484 rest datasets, NeuroImage, 2012

62 Parametric vs non-parametric statistics Non-parametric algorithms (e.g. a permutation test) can sometimes provide more reliable results Empirically estimate the null distribution, by analyzing data similar to the original dataset Obtain similar data by randomly permuting volumes

63 Eklund et al., Does parametric fmri analysis with SPM yield valid results? An empirical study of 1484 rest datasets, NeuroImage, 2012

64 Eklund et al., Does parametric fmri analysis with SPM yield valid results? An empirical study of 1484 rest datasets, NeuroImage, 2012

65 Demo 4, permutation based first level analysis in BROCCOLI

66 Demo 4, Summary Can easily add a permutation test by using the option -permute Gives permutation based p-values, corrected for multiple comparisons 1,000 permutations in about 6 seconds

67 What about Bayesian statistics?

68 Bayesian fmri analysis Most Bayesian fmri analyses rely on variational Bayes to obtain an analytical expression of the posterior distribution Variational Bayes is based on a number of assumptions and approximations For this reason, BROCCOLI instead uses MCMC, runs many MCMC chains in parallel Ferreira da Silva., cudabayesreg: Parallel implementation of a Bayesian multilevel model for fmri data analysis, Journal of statistical software, 44, 2011

69 Demo 5, Bayesian first level analysis in BROCCOLI

70 Demo 5, Summary Can easily do MCMC based Bayesian analysis by using the option -bayesian Currently only supports 2 regressors and AR(1) Currently only supports first level analysis 1,000 MCMC iterations in about 20 seconds (straight forward Matlab implementation: 2.5 h)

71 Permutation based second level analysis in BROCCOLI Very similar to randomise in FSL Supports voxel level, cluster extent and cluster mass inference Threshold free cluster enhancement being implemented

72 RandomiseGroupLevel - Bash RandomiseGroupLevel volumes.nii [options] Some of the options are -platform -device -groupmean -design -contrasts -inferencemode -mask OpenCL platform to use OpenCL device to use One sample t-test Regressors Contrasts Voxel, cluster extent, cluster mass Restrict analysis to mask

73 Demo 6, permutation based second level analysis in BROCCOLI

74 Demo 6, Summary Usage very similar to randomise in FSL BROCCOLI gives the same results as randomise for one regressor, not yet for several regressors BROCCOLI is ~85 times faster for 1 regressor, ~250 times faster for 25 regressors Lower speedup when compared to the parallel version of randomise

75 What about the future?

76 Future extensions of BROCCOLI Permutation based group level analysis using a method similar to FLAME in FSL or 3dMEMA in AFNI (includes the variance of each subject)

77 Future extensions of BROCCOLI Statistical analysis using canonical correlation analysis (CCA) Combines neighbouring voxels or filter responses in an optimal way, can give a higher statistical power Use permutation test to estimate complicated null distribution Friman et al., Adaptive analysis of fmri data, NeuroImage, 2003

78 Future extensions of BROCCOLI Support vector machines (SVM) Permutation based SVM, to get p-values of the classifier weights SVM relies on matrix operations, CUBLAS for CUDA, clblas for OpenCL Catanzaro et al., Fast support vector machine training and classification on graphics processors, ICML, , 2008

79 Future extensions of BROCCOLI Independent component analysis (ICA) Permutation based ICA, to get p-values of time series and spatial maps ICA can be performed using matrix operations, CUBLAS for CUDA, clblas for OpenCL Raimondo et al., CUDAICA: GPU optimization of infomax-ica EEG Analysis, Computational Intelligence and Neuroscience, 2012

80 Thanks to Stephen LaConte Postdoc supervisor Paul Dufort GPU guru Mattias Villani Bayesian guru Daniel Forsberg Image registration guru Cyrus Eierud Beta tester Morgan Hough Python guru Satrajit Ghosh Python guru Miha Cancula Python guru Hans Knutsson Phd supervisor Mats Andersson Phd supervisor

81 Questions or comments? Beta testers needed!

EMPIRICALLY INVESTIGATING THE STATISTICAL VALIDITY OF SPM, FSL AND AFNI FOR SINGLE SUBJECT FMRI ANALYSIS

EMPIRICALLY INVESTIGATING THE STATISTICAL VALIDITY OF SPM, FSL AND AFNI FOR SINGLE SUBJECT FMRI ANALYSIS EMPIRICALLY INVESTIGATING THE STATISTICAL VALIDITY OF SPM, FSL AND AFNI FOR SINGLE SUBJECT FMRI ANALYSIS Anders Eklund a,b,c, Thomas Nichols d, Mats Andersson a,c, Hans Knutsson a,c a Department of Biomedical

More information

Cluster failure: Why fmri inferences for spatial extent have inflated false positive rates

Cluster failure: Why fmri inferences for spatial extent have inflated false positive rates Supporting Information Appendix Cluster failure: Why fmri inferences for spatial extent have inflated false positive rates Anders Eklund, Thomas Nichols, Hans Knutsson Methods Resting state fmri data Resting

More information

Multivariate fmri Analysis using Canonical Correlation Analysis instead of Classifiers, Comment on Todd et al.

Multivariate fmri Analysis using Canonical Correlation Analysis instead of Classifiers, Comment on Todd et al. Multivariate fmri Analysis using Canonical Correlation Analysis instead of Classifiers, Comment on Todd et al. Anders Eklund a, Hans Knutsson bc a Virginia Tech Carilion Research Institute, Virginia Tech,

More information

Improving CCA based fmri Analysis by Covariance Pooling - Using the GPU for Statistical Inference

Improving CCA based fmri Analysis by Covariance Pooling - Using the GPU for Statistical Inference Improving CCA based fmri Analysis by Covariance Pooling - Using the GPU for Statistical Inference Anders Eklund, Mats Andersson and Hans Knutsson Linköping University Post Print N.B.: When citing this

More information

fmri Analysis on the GPU - Possibilities and Challenges

fmri Analysis on the GPU - Possibilities and Challenges fmri Analysis on the GPU - Possibilities and Challenges Anders Eklund, Mats Andersson and Hans Knutsson Linköping University Post Print N.B.: When citing this work, cite the original article. Original

More information

A GPU Accelerated Interactive Interface for Exploratory Functional Connectivity Analysis of fmri Data

A GPU Accelerated Interactive Interface for Exploratory Functional Connectivity Analysis of fmri Data A GPU Accelerated Interactive Interface for Exploratory Functional Connectivity Analysis of fmri Data Anders Eklund, Ola Friman, Mats Andersson and Hans Knutsson Linköping University Post Print N.B.: When

More information

7/15/2016 ARE YOUR ANALYSES TOO WHY IS YOUR ANALYSIS PARAMETRIC? PARAMETRIC? That s not Normal!

7/15/2016 ARE YOUR ANALYSES TOO WHY IS YOUR ANALYSIS PARAMETRIC? PARAMETRIC? That s not Normal! ARE YOUR ANALYSES TOO PARAMETRIC? That s not Normal! Martin M Monti http://montilab.psych.ucla.edu WHY IS YOUR ANALYSIS PARAMETRIC? i. Optimal power (defined as the probability to detect a real difference)

More information

Computational Neuroanatomy

Computational Neuroanatomy Computational Neuroanatomy John Ashburner john@fil.ion.ucl.ac.uk Smoothing Motion Correction Between Modality Co-registration Spatial Normalisation Segmentation Morphometry Overview fmri time-series kernel

More information

A Functional Connectivity Inspired Approach to Non-Local fmri Analysis

A Functional Connectivity Inspired Approach to Non-Local fmri Analysis A Functional Connectivity Inspired Approach to Non-Local fmri Analysis Anders Eklund, Mats Andersson and Hans Knutsson Linköping University Post Print N.B.: When citing this work, cite the original article.

More information

An Introduction To Automatic Tissue Classification Of Brain MRI. Colm Elliott Mar 2014

An Introduction To Automatic Tissue Classification Of Brain MRI. Colm Elliott Mar 2014 An Introduction To Automatic Tissue Classification Of Brain MRI Colm Elliott Mar 2014 Tissue Classification Tissue classification is part of many processing pipelines. We often want to classify each voxel

More information

Statistical Analysis of Neuroimaging Data. Phebe Kemmer BIOS 516 Sept 24, 2015

Statistical Analysis of Neuroimaging Data. Phebe Kemmer BIOS 516 Sept 24, 2015 Statistical Analysis of Neuroimaging Data Phebe Kemmer BIOS 516 Sept 24, 2015 Review from last time Structural Imaging modalities MRI, CAT, DTI (diffusion tensor imaging) Functional Imaging modalities

More information

Functional MRI in Clinical Research and Practice Preprocessing

Functional MRI in Clinical Research and Practice Preprocessing Functional MRI in Clinical Research and Practice Preprocessing fmri Preprocessing Slice timing correction Geometric distortion correction Head motion correction Temporal filtering Intensity normalization

More information

Basic fmri Design and Analysis. Preprocessing

Basic fmri Design and Analysis. Preprocessing Basic fmri Design and Analysis Preprocessing fmri Preprocessing Slice timing correction Geometric distortion correction Head motion correction Temporal filtering Intensity normalization Spatial filtering

More information

SPM8 for Basic and Clinical Investigators. Preprocessing. fmri Preprocessing

SPM8 for Basic and Clinical Investigators. Preprocessing. fmri Preprocessing SPM8 for Basic and Clinical Investigators Preprocessing fmri Preprocessing Slice timing correction Geometric distortion correction Head motion correction Temporal filtering Intensity normalization Spatial

More information

Analysis of Functional MRI Timeseries Data Using Signal Processing Techniques

Analysis of Functional MRI Timeseries Data Using Signal Processing Techniques Analysis of Functional MRI Timeseries Data Using Signal Processing Techniques Sea Chen Department of Biomedical Engineering Advisors: Dr. Charles A. Bouman and Dr. Mark J. Lowe S. Chen Final Exam October

More information

EPI Data Are Acquired Serially. EPI Data Are Acquired Serially 10/23/2011. Functional Connectivity Preprocessing. fmri Preprocessing

EPI Data Are Acquired Serially. EPI Data Are Acquired Serially 10/23/2011. Functional Connectivity Preprocessing. fmri Preprocessing Functional Connectivity Preprocessing Geometric distortion Head motion Geometric distortion Head motion EPI Data Are Acquired Serially EPI Data Are Acquired Serially descending 1 EPI Data Are Acquired

More information

MULTIVARIATE ANALYSES WITH fmri DATA

MULTIVARIATE ANALYSES WITH fmri DATA MULTIVARIATE ANALYSES WITH fmri DATA Sudhir Shankar Raman Translational Neuromodeling Unit (TNU) Institute for Biomedical Engineering University of Zurich & ETH Zurich Motivation Modelling Concepts Learning

More information

Journal of Articles in Support of The Null Hypothesis

Journal of Articles in Support of The Null Hypothesis Data Preprocessing Martin M. Monti, PhD UCLA Psychology NITP 2016 Typical (task-based) fmri analysis sequence Image Pre-processing Single Subject Analysis Group Analysis Journal of Articles in Support

More information

Function-Structure Integration in FreeSurfer

Function-Structure Integration in FreeSurfer Function-Structure Integration in FreeSurfer Outline Function-Structure Integration Function-Structure Registration in FreeSurfer fmri Analysis Preprocessing First-Level Analysis Higher-Level (Group) Analysis

More information

Lecture 1: Introduction and Computational Thinking

Lecture 1: Introduction and Computational Thinking PASI Summer School Advanced Algorithmic Techniques for GPUs Lecture 1: Introduction and Computational Thinking 1 Course Objective To master the most commonly used algorithm techniques and computational

More information

FMRI data: Independent Component Analysis (GIFT) & Connectivity Analysis (FNC)

FMRI data: Independent Component Analysis (GIFT) & Connectivity Analysis (FNC) FMRI data: Independent Component Analysis (GIFT) & Connectivity Analysis (FNC) Software: Matlab Toolbox: GIFT & FNC Yingying Wang, Ph.D. in Biomedical Engineering 10 16 th, 2014 PI: Dr. Nadine Gaab Outline

More information

GPU Programming Using NVIDIA CUDA

GPU Programming Using NVIDIA CUDA GPU Programming Using NVIDIA CUDA Siddhante Nangla 1, Professor Chetna Achar 2 1, 2 MET s Institute of Computer Science, Bandra Mumbai University Abstract: GPGPU or General-Purpose Computing on Graphics

More information

Supporting Data Parallelism in Matcloud: Final Report

Supporting Data Parallelism in Matcloud: Final Report Supporting Data Parallelism in Matcloud: Final Report Yongpeng Zhang, Xing Wu 1 Overview Matcloud is an on-line service to run Matlab-like script on client s web browser. Internally it is accelerated by

More information

CS 179 Lecture 16. Logistic Regression & Parallel SGD

CS 179 Lecture 16. Logistic Regression & Parallel SGD CS 179 Lecture 16 Logistic Regression & Parallel SGD 1 Outline logistic regression (stochastic) gradient descent parallelizing SGD for neural nets (with emphasis on Google s distributed neural net implementation)

More information

Introduction to Neuroimaging Janaina Mourao-Miranda

Introduction to Neuroimaging Janaina Mourao-Miranda Introduction to Neuroimaging Janaina Mourao-Miranda Neuroimaging techniques have changed the way neuroscientists address questions about functional anatomy, especially in relation to behavior and clinical

More information

Bayesian Inference in fmri Will Penny

Bayesian Inference in fmri Will Penny Bayesian Inference in fmri Will Penny Bayesian Approaches in Neuroscience Karolinska Institutet, Stockholm February 2016 Overview Posterior Probability Maps Hemodynamic Response Functions Population

More information

Fmri Spatial Processing

Fmri Spatial Processing Educational Course: Fmri Spatial Processing Ray Razlighi Jun. 8, 2014 Spatial Processing Spatial Re-alignment Geometric distortion correction Spatial Normalization Smoothing Why, When, How, Which Why is

More information

DEEP NEURAL NETWORKS AND GPUS. Julie Bernauer

DEEP NEURAL NETWORKS AND GPUS. Julie Bernauer DEEP NEURAL NETWORKS AND GPUS Julie Bernauer GPU Computing GPU Computing Run Computations on GPUs x86 CUDA Framework to Program NVIDIA GPUs A simple sum of two vectors (arrays) in C void vector_add(int

More information

Autonomous Navigation for Flying Robots

Autonomous Navigation for Flying Robots Computer Vision Group Prof. Daniel Cremers Autonomous Navigation for Flying Robots Lecture 7.1: 2D Motion Estimation in Images Jürgen Sturm Technische Universität München 3D to 2D Perspective Projections

More information

PLB-HeC: A Profile-based Load-Balancing Algorithm for Heterogeneous CPU-GPU Clusters

PLB-HeC: A Profile-based Load-Balancing Algorithm for Heterogeneous CPU-GPU Clusters PLB-HeC: A Profile-based Load-Balancing Algorithm for Heterogeneous CPU-GPU Clusters IEEE CLUSTER 2015 Chicago, IL, USA Luis Sant Ana 1, Daniel Cordeiro 2, Raphael Camargo 1 1 Federal University of ABC,

More information

AFNI Preprocessing: Outline, Recommendations, and New(ish) Stuff. Robert W Cox SSCC / NIMH & NINDS / NIH / DHHS / USA / EARTH

AFNI Preprocessing: Outline, Recommendations, and New(ish) Stuff. Robert W Cox SSCC / NIMH & NINDS / NIH / DHHS / USA / EARTH AFNI Preprocessing: Outline, Recommendations, and New(ish) Stuff Robert W Cox SSCC / NIMH & NINDS / NIH / DHHS / USA / EARTH HBM 2016 As a work of a US Government official, this presentation is not copyrighted

More information

Pattern Recognition for Neuroimaging Data

Pattern Recognition for Neuroimaging Data Pattern Recognition for Neuroimaging Data Edinburgh, SPM course April 2013 C. Phillips, Cyclotron Research Centre, ULg, Belgium http://www.cyclotron.ulg.ac.be Overview Introduction Univariate & multivariate

More information

Correction for multiple comparisons. Cyril Pernet, PhD SBIRC/SINAPSE University of Edinburgh

Correction for multiple comparisons. Cyril Pernet, PhD SBIRC/SINAPSE University of Edinburgh Correction for multiple comparisons Cyril Pernet, PhD SBIRC/SINAPSE University of Edinburgh Overview Multiple comparisons correction procedures Levels of inferences (set, cluster, voxel) Circularity issues

More information

BHMSMAfMRI User Manual

BHMSMAfMRI User Manual BHMSMAfMRI User Manual Nilotpal Sanyal Bayesian and Interdisciplinary Research Unit Indian Statistical Institute, Calcutta, India nsanyal v@isical.ac.in Welcome to BHMSMAfMRI, an R package to analyze functional

More information

First-level fmri modeling

First-level fmri modeling First-level fmri modeling Monday, Lecture 3 Jeanette Mumford University of Wisconsin - Madison What do we need to remember from the last lecture? What is the general structure of a t- statistic? How about

More information

William Yang Group 14 Mentor: Dr. Rogerio Richa Visual Tracking of Surgical Tools in Retinal Surgery using Particle Filtering

William Yang Group 14 Mentor: Dr. Rogerio Richa Visual Tracking of Surgical Tools in Retinal Surgery using Particle Filtering Mutual Information Computation and Maximization Using GPU Yuping Lin and Gérard Medioni Computer Vision and Pattern Recognition Workshops (CVPR) Anchorage, AK, pp. 1-6, June 2008 Project Summary and Paper

More information

Using Existing Numerical Libraries on Spark

Using Existing Numerical Libraries on Spark Using Existing Numerical Libraries on Spark Brian Spector Chicago Spark Users Meetup June 24 th, 2015 Experts in numerical algorithms and HPC services How to use existing libraries on Spark Call algorithm

More information

Multivariate pattern classification

Multivariate pattern classification Multivariate pattern classification Thomas Wolbers Space & Ageing Laboratory (www.sal.mvm.ed.ac.uk) Centre for Cognitive and Neural Systems & Centre for Cognitive Ageing and Cognitive Epidemiology Outline

More information

Using a GPU in InSAR processing to improve performance

Using a GPU in InSAR processing to improve performance Using a GPU in InSAR processing to improve performance Rob Mellors, ALOS PI 152 San Diego State University David Sandwell University of California, San Diego What is a GPU? (Graphic Processor Unit) A graphics

More information

Representational similarity analysis. Dr Ian Charest,

Representational similarity analysis. Dr Ian Charest, Representational similarity analysis Dr Ian Charest, Edinburgh, April 219 A space for neuroimaging studies Pattern across subjects Cross-subject correlation studies (e.g. social neuroscience) Raizada et

More information

Exploiting the OpenPOWER Platform for Big Data Analytics and Cognitive. Rajesh Bordawekar and Ruchir Puri IBM T. J. Watson Research Center

Exploiting the OpenPOWER Platform for Big Data Analytics and Cognitive. Rajesh Bordawekar and Ruchir Puri IBM T. J. Watson Research Center Exploiting the OpenPOWER Platform for Big Data Analytics and Cognitive Rajesh Bordawekar and Ruchir Puri IBM T. J. Watson Research Center 3/17/2015 2014 IBM Corporation Outline IBM OpenPower Platform Accelerating

More information

Large Displacement Optical Flow & Applications

Large Displacement Optical Flow & Applications Large Displacement Optical Flow & Applications Narayanan Sundaram, Kurt Keutzer (Parlab) In collaboration with Thomas Brox (University of Freiburg) Michael Tao (University of California Berkeley) Parlab

More information

Group (Level 2) fmri Data Analysis - Lab 4

Group (Level 2) fmri Data Analysis - Lab 4 Group (Level 2) fmri Data Analysis - Lab 4 Index Goals of this Lab Before Getting Started The Chosen Ten Checking Data Quality Create a Mean Anatomical of the Group Group Analysis: One-Sample T-Test Examine

More information

SPM8 for Basic and Clinical Investigators. Preprocessing

SPM8 for Basic and Clinical Investigators. Preprocessing SPM8 for Basic and Clinical Investigators Preprocessing fmri Preprocessing Slice timing correction Geometric distortion correction Head motion correction Temporal filtering Intensity normalization Spatial

More information

FSL Pre-Processing Pipeline

FSL Pre-Processing Pipeline The Art and Pitfalls of fmri Preprocessing FSL Pre-Processing Pipeline Mark Jenkinson FMRIB Centre, University of Oxford FSL Pre-Processing Pipeline Standard pre-processing: Task fmri Resting-state fmri

More information

COSC160: Detection and Classification. Jeremy Bolton, PhD Assistant Teaching Professor

COSC160: Detection and Classification. Jeremy Bolton, PhD Assistant Teaching Professor COSC160: Detection and Classification Jeremy Bolton, PhD Assistant Teaching Professor Outline I. Problem I. Strategies II. Features for training III. Using spatial information? IV. Reducing dimensionality

More information

Analysis of fmri data within Brainvisa Example with the Saccades database

Analysis of fmri data within Brainvisa Example with the Saccades database Analysis of fmri data within Brainvisa Example with the Saccades database 18/11/2009 Note : All the sentences in italic correspond to informations relative to the specific dataset under study TP participants

More information

CS 229 Final Project Report Learning to Decode Cognitive States of Rat using Functional Magnetic Resonance Imaging Time Series

CS 229 Final Project Report Learning to Decode Cognitive States of Rat using Functional Magnetic Resonance Imaging Time Series CS 229 Final Project Report Learning to Decode Cognitive States of Rat using Functional Magnetic Resonance Imaging Time Series Jingyuan Chen //Department of Electrical Engineering, cjy2010@stanford.edu//

More information

SPM Introduction. SPM : Overview. SPM: Preprocessing SPM! SPM: Preprocessing. Scott Peltier. FMRI Laboratory University of Michigan

SPM Introduction. SPM : Overview. SPM: Preprocessing SPM! SPM: Preprocessing. Scott Peltier. FMRI Laboratory University of Michigan SPM Introduction Scott Peltier FMRI Laboratory University of Michigan! Slides adapted from T. Nichols SPM! SPM : Overview Library of MATLAB and C functions Graphical user interface Four main components:

More information

Accelerated Machine Learning Algorithms in Python

Accelerated Machine Learning Algorithms in Python Accelerated Machine Learning Algorithms in Python Patrick Reilly, Leiming Yu, David Kaeli reilly.pa@husky.neu.edu Northeastern University Computer Architecture Research Lab Outline Motivation and Goals

More information

Quality Checking an fmri Group Result (art_groupcheck)

Quality Checking an fmri Group Result (art_groupcheck) Quality Checking an fmri Group Result (art_groupcheck) Paul Mazaika, Feb. 24, 2009 A statistical parameter map of fmri group analyses relies on the assumptions of the General Linear Model (GLM). The assumptions

More information

FSL Pre-Processing Pipeline

FSL Pre-Processing Pipeline The Art and Pitfalls of fmri Preprocessing FSL Pre-Processing Pipeline Mark Jenkinson FMRIB Centre, University of Oxford FSL Pre-Processing Pipeline Standard pre-processing: Task fmri Resting-state fmri

More information

SPM Introduction SPM! Scott Peltier. FMRI Laboratory University of Michigan. Software to perform computation, manipulation and display of imaging data

SPM Introduction SPM! Scott Peltier. FMRI Laboratory University of Michigan. Software to perform computation, manipulation and display of imaging data SPM Introduction Scott Peltier FMRI Laboratory University of Michigan Slides adapted from T. Nichols SPM! Software to perform computation, manipulation and display of imaging data 1 1 SPM : Overview Library

More information

Group Sta*s*cs in MEG/EEG

Group Sta*s*cs in MEG/EEG Group Sta*s*cs in MEG/EEG Will Woods NIF Fellow Brain and Psychological Sciences Research Centre Swinburne University of Technology A Cau*onary tale. A Cau*onary tale. A Cau*onary tale. Overview Introduc*on

More information

HPC Middle East. KFUPM HPC Workshop April Mohamed Mekias HPC Solutions Consultant. Introduction to CUDA programming

HPC Middle East. KFUPM HPC Workshop April Mohamed Mekias HPC Solutions Consultant. Introduction to CUDA programming KFUPM HPC Workshop April 29-30 2015 Mohamed Mekias HPC Solutions Consultant Introduction to CUDA programming 1 Agenda GPU Architecture Overview Tools of the Trade Introduction to CUDA C Patterns of Parallel

More information

Preprocessing of fmri data

Preprocessing of fmri data Preprocessing of fmri data Pierre Bellec CRIUGM, DIRO, UdM Flowchart of the NIAK fmri preprocessing pipeline fmri run 1 fmri run N individual datasets CIVET NUC, segmentation, spatial normalization slice

More information

Chapter 3 Image Registration. Chapter 3 Image Registration

Chapter 3 Image Registration. Chapter 3 Image Registration Chapter 3 Image Registration Distributed Algorithms for Introduction (1) Definition: Image Registration Input: 2 images of the same scene but taken from different perspectives Goal: Identify transformation

More information

Bayesian Spatiotemporal Modeling with Hierarchical Spatial Priors for fmri

Bayesian Spatiotemporal Modeling with Hierarchical Spatial Priors for fmri Bayesian Spatiotemporal Modeling with Hierarchical Spatial Priors for fmri Galin L. Jones 1 School of Statistics University of Minnesota March 2015 1 Joint with Martin Bezener and John Hughes Experiment

More information

Large scale Imaging on Current Many- Core Platforms

Large scale Imaging on Current Many- Core Platforms Large scale Imaging on Current Many- Core Platforms SIAM Conf. on Imaging Science 2012 May 20, 2012 Dr. Harald Köstler Chair for System Simulation Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen,

More information

FSL Workshop Session 3 David Smith & John Clithero

FSL Workshop Session 3 David Smith & John Clithero FSL Workshop 12.09.08 Session 3 David Smith & John Clithero What is MELODIC? Probabilistic ICA Improves upon standard ICA Allows for inference Avoids over-fitting Three stage process ( ppca ) 1.) Dimension

More information

Applying Supervised Learning

Applying Supervised Learning Applying Supervised Learning When to Consider Supervised Learning A supervised learning algorithm takes a known set of input data (the training set) and known responses to the data (output), and trains

More information

Research Article True 4D Image Denoising on the GPU

Research Article True 4D Image Denoising on the GPU International Journal of Biomedical Imaging Volume 2011, Article ID 952819, 16 pages doi:10.1155/2011/952819 Research Article True 4D Image Denoising on the GPU Anders Eklund, 1, 2 Mats Andersson, 1, 2

More information

LOCAL-GLOBAL OPTICAL FLOW FOR IMAGE REGISTRATION

LOCAL-GLOBAL OPTICAL FLOW FOR IMAGE REGISTRATION LOCAL-GLOBAL OPTICAL FLOW FOR IMAGE REGISTRATION Ammar Zayouna Richard Comley Daming Shi Middlesex University School of Engineering and Information Sciences Middlesex University, London NW4 4BT, UK A.Zayouna@mdx.ac.uk

More information

EE795: Computer Vision and Intelligent Systems

EE795: Computer Vision and Intelligent Systems EE795: Computer Vision and Intelligent Systems Spring 2012 TTh 17:30-18:45 FDH 204 Lecture 10 130221 http://www.ee.unlv.edu/~b1morris/ecg795/ 2 Outline Review Canny Edge Detector Hough Transform Feature-Based

More information

Multivariate Pattern Classification. Thomas Wolbers Space and Aging Laboratory Centre for Cognitive and Neural Systems

Multivariate Pattern Classification. Thomas Wolbers Space and Aging Laboratory Centre for Cognitive and Neural Systems Multivariate Pattern Classification Thomas Wolbers Space and Aging Laboratory Centre for Cognitive and Neural Systems Outline WHY PATTERN CLASSIFICATION? PROCESSING STREAM PREPROCESSING / FEATURE REDUCTION

More information

Large-scale Deep Unsupervised Learning using Graphics Processors

Large-scale Deep Unsupervised Learning using Graphics Processors Large-scale Deep Unsupervised Learning using Graphics Processors Rajat Raina Anand Madhavan Andrew Y. Ng Stanford University Learning from unlabeled data Classify vs. car motorcycle Input space Unlabeled

More information

MRI Segmentation. MRI Bootcamp, 14 th of January J. Miguel Valverde

MRI Segmentation. MRI Bootcamp, 14 th of January J. Miguel Valverde MRI Segmentation MRI Bootcamp, 14 th of January 2019 Segmentation Segmentation Information Segmentation Algorithms Approach Types of Information Local 4 45 100 110 80 50 76 42 27 186 177 120 167 111 56

More information

The organization of the human cerebral cortex estimated by intrinsic functional connectivity

The organization of the human cerebral cortex estimated by intrinsic functional connectivity 1 The organization of the human cerebral cortex estimated by intrinsic functional connectivity Journal: Journal of Neurophysiology Author: B. T. Thomas Yeo, et al Link: https://www.ncbi.nlm.nih.gov/pubmed/21653723

More information

Phase-Based Non-Rigid 3D Image Registration: From Minutes to Seconds Using CUDA

Phase-Based Non-Rigid 3D Image Registration: From Minutes to Seconds Using CUDA Phase-Based Non-Rigid 3D Image Registration: From Minutes to Seconds Using CUDA Daniel Forsberg 1,2,3, Anders Eklund 1,2, Mats Andersson 1,2, and Hans Knutsson 1,2 1 Department of Biomedical Engineering,

More information

Basic principles of MR image analysis. Basic principles of MR image analysis. Basic principles of MR image analysis

Basic principles of MR image analysis. Basic principles of MR image analysis. Basic principles of MR image analysis Basic principles of MR image analysis Basic principles of MR image analysis Julien Milles Leiden University Medical Center Terminology of fmri Brain extraction Registration Linear registration Non-linear

More information

Deep Learning. Volker Tresp Summer 2014

Deep Learning. Volker Tresp Summer 2014 Deep Learning Volker Tresp Summer 2014 1 Neural Network Winter and Revival While Machine Learning was flourishing, there was a Neural Network winter (late 1990 s until late 2000 s) Around 2010 there

More information

Introduction to GPU Computing. 周国峰 Wuhan University 2017/10/13

Introduction to GPU Computing. 周国峰 Wuhan University 2017/10/13 Introduction to GPU Computing chandlerz@nvidia.com 周国峰 Wuhan University 2017/10/13 GPU and Its Application 3 Ways to Develop Your GPU APP An Example to Show the Developments Add GPUs: Accelerate Science

More information

CS 565 Computer Vision. Nazar Khan PUCIT Lectures 15 and 16: Optic Flow

CS 565 Computer Vision. Nazar Khan PUCIT Lectures 15 and 16: Optic Flow CS 565 Computer Vision Nazar Khan PUCIT Lectures 15 and 16: Optic Flow Introduction Basic Problem given: image sequence f(x, y, z), where (x, y) specifies the location and z denotes time wanted: displacement

More information

GPU ACCELERATED DATABASE MANAGEMENT SYSTEMS

GPU ACCELERATED DATABASE MANAGEMENT SYSTEMS CIS 601 - Graduate Seminar Presentation 1 GPU ACCELERATED DATABASE MANAGEMENT SYSTEMS PRESENTED BY HARINATH AMASA CSU ID: 2697292 What we will talk about.. Current problems GPU What are GPU Databases GPU

More information

1 Introduction Motivation and Aims Functional Imaging Computational Neuroanatomy... 12

1 Introduction Motivation and Aims Functional Imaging Computational Neuroanatomy... 12 Contents 1 Introduction 10 1.1 Motivation and Aims....... 10 1.1.1 Functional Imaging.... 10 1.1.2 Computational Neuroanatomy... 12 1.2 Overview of Chapters... 14 2 Rigid Body Registration 18 2.1 Introduction.....

More information

Automatic Registration-Based Segmentation for Neonatal Brains Using ANTs and Atropos

Automatic Registration-Based Segmentation for Neonatal Brains Using ANTs and Atropos Automatic Registration-Based Segmentation for Neonatal Brains Using ANTs and Atropos Jue Wu and Brian Avants Penn Image Computing and Science Lab, University of Pennsylvania, Philadelphia, USA Abstract.

More information

Independent Component Analysis of fmri Data

Independent Component Analysis of fmri Data Independent Component Analysis of fmri Data Denise Miller April 2005 Introduction Techniques employed to analyze functional magnetic resonance imaging (fmri) data typically use some form of univariate

More information

Introduction to fmri. Pre-processing

Introduction to fmri. Pre-processing Introduction to fmri Pre-processing Tibor Auer Department of Psychology Research Fellow in MRI Data Types Anatomical data: T 1 -weighted, 3D, 1/subject or session - (ME)MPRAGE/FLASH sequence, undistorted

More information

Exploiting GPU Caches in Sparse Matrix Vector Multiplication. Yusuke Nagasaka Tokyo Institute of Technology

Exploiting GPU Caches in Sparse Matrix Vector Multiplication. Yusuke Nagasaka Tokyo Institute of Technology Exploiting GPU Caches in Sparse Matrix Vector Multiplication Yusuke Nagasaka Tokyo Institute of Technology Sparse Matrix Generated by FEM, being as the graph data Often require solving sparse linear equation

More information

Templates, Image Pyramids, and Filter Banks

Templates, Image Pyramids, and Filter Banks Templates, Image Pyramids, and Filter Banks Computer Vision James Hays, Brown Slides: Hoiem and others Reminder Project due Friday Fourier Bases Teases away fast vs. slow changes in the image. This change

More information

Accelerating image registration on GPUs

Accelerating image registration on GPUs Accelerating image registration on GPUs Harald Köstler, Sunil Ramgopal Tatavarty SIAM Conference on Imaging Science (IS10) 13.4.2010 Contents Motivation: Image registration with FAIR GPU Programming Combining

More information

Acceleration of Probabilistic Tractography Using Multi-GPU Parallel Processing. Jungsoo Lee, Sun Mi Park, Dae-Shik Kim

Acceleration of Probabilistic Tractography Using Multi-GPU Parallel Processing. Jungsoo Lee, Sun Mi Park, Dae-Shik Kim Acceleration of Probabilistic Tractography Using Multi-GPU Parallel Processing Jungsoo Lee, Sun Mi Park, Dae-Shik Kim Introduction In particular, probabilistic tractography requires relatively long computation

More information

Applied Bayesian Nonparametrics 5. Spatial Models via Gaussian Processes, not MRFs Tutorial at CVPR 2012 Erik Sudderth Brown University

Applied Bayesian Nonparametrics 5. Spatial Models via Gaussian Processes, not MRFs Tutorial at CVPR 2012 Erik Sudderth Brown University Applied Bayesian Nonparametrics 5. Spatial Models via Gaussian Processes, not MRFs Tutorial at CVPR 2012 Erik Sudderth Brown University NIPS 2008: E. Sudderth & M. Jordan, Shared Segmentation of Natural

More information

Neuroimaging and mathematical modelling Lesson 2: Voxel Based Morphometry

Neuroimaging and mathematical modelling Lesson 2: Voxel Based Morphometry Neuroimaging and mathematical modelling Lesson 2: Voxel Based Morphometry Nivedita Agarwal, MD Nivedita.agarwal@apss.tn.it Nivedita.agarwal@unitn.it Volume and surface morphometry Brain volume White matter

More information

Direct Matrix Factorization and Alignment Refinement: Application to Defect Detection

Direct Matrix Factorization and Alignment Refinement: Application to Defect Detection Direct Matrix Factorization and Alignment Refinement: Application to Defect Detection Zhen Qin (University of California, Riverside) Peter van Beek & Xu Chen (SHARP Labs of America, Camas, WA) 2015/8/30

More information

White Pixel Artifact. Caused by a noise spike during acquisition Spike in K-space <--> sinusoid in image space

White Pixel Artifact. Caused by a noise spike during acquisition Spike in K-space <--> sinusoid in image space White Pixel Artifact Caused by a noise spike during acquisition Spike in K-space sinusoid in image space Susceptibility Artifacts Off-resonance artifacts caused by adjacent regions with different

More information

Photo-realistic Renderings for Machines Seong-heum Kim

Photo-realistic Renderings for Machines Seong-heum Kim Photo-realistic Renderings for Machines 20105034 Seong-heum Kim CS580 Student Presentations 2016.04.28 Photo-realistic Renderings for Machines Scene radiances Model descriptions (Light, Shape, Material,

More information

The Shogun Machine Learning Toolbox

The Shogun Machine Learning Toolbox The Shogun Machine Learning Toolbox heiko.strathmann@gmail.com Europython 2014 July 24, 2014 Outline Overview Machine Learning Features Technical Features Community A bit about me Heiko Strathmann - http://herrstrathmann.de/

More information

High-Performance Computing Using GPUs

High-Performance Computing Using GPUs High-Performance Computing Using GPUs Luca Caucci caucci@email.arizona.edu Center for Gamma-Ray Imaging November 7, 2012 Outline Slide 1 of 27 Why GPUs? What is CUDA? The CUDA programming model Anatomy

More information

Object detection using Region Proposals (RCNN) Ernest Cheung COMP Presentation

Object detection using Region Proposals (RCNN) Ernest Cheung COMP Presentation Object detection using Region Proposals (RCNN) Ernest Cheung COMP790-125 Presentation 1 2 Problem to solve Object detection Input: Image Output: Bounding box of the object 3 Object detection using CNN

More information

Basic Introduction to Data Analysis. Block Design Demonstration. Robert Savoy

Basic Introduction to Data Analysis. Block Design Demonstration. Robert Savoy Basic Introduction to Data Analysis Block Design Demonstration Robert Savoy Sample Block Design Experiment Demonstration Use of Visual and Motor Task Separability of Responses Combined Visual and Motor

More information

Dense Image-based Motion Estimation Algorithms & Optical Flow

Dense Image-based Motion Estimation Algorithms & Optical Flow Dense mage-based Motion Estimation Algorithms & Optical Flow Video A video is a sequence of frames captured at different times The video data is a function of v time (t) v space (x,y) ntroduction to motion

More information

Topics to be Covered in the Rest of the Semester. CSci 4968 and 6270 Computational Vision Lecture 15 Overview of Remainder of the Semester

Topics to be Covered in the Rest of the Semester. CSci 4968 and 6270 Computational Vision Lecture 15 Overview of Remainder of the Semester Topics to be Covered in the Rest of the Semester CSci 4968 and 6270 Computational Vision Lecture 15 Overview of Remainder of the Semester Charles Stewart Department of Computer Science Rensselaer Polytechnic

More information

Graphical Models, Bayesian Method, Sampling, and Variational Inference

Graphical Models, Bayesian Method, Sampling, and Variational Inference Graphical Models, Bayesian Method, Sampling, and Variational Inference With Application in Function MRI Analysis and Other Imaging Problems Wei Liu Scientific Computing and Imaging Institute University

More information

GPU-PCC: A GPU Based Technique to Compute Pairwise Pearson s Correlation Coefficients for Big fmri Data

GPU-PCC: A GPU Based Technique to Compute Pairwise Pearson s Correlation Coefficients for Big fmri Data Western Michigan University ScholarWorks at WMU Parallel Computing and Data Science Lab Technical Reports Computer Science 2017 GPU-PCC: A GPU Based Technique to Compute Pairwise Pearson s Correlation

More information

Extending the GLM. Outline. Mixed effects motivation Evaluating mixed effects methods Three methods. Conclusions. Overview

Extending the GLM. Outline. Mixed effects motivation Evaluating mixed effects methods Three methods. Conclusions. Overview Extending the GLM So far, we have considered the GLM for one run in one subject The same logic can be applied to multiple runs and multiple subjects GLM Stats For any given region, we can evaluate the

More information

ELEC Dr Reji Mathew Electrical Engineering UNSW

ELEC Dr Reji Mathew Electrical Engineering UNSW ELEC 4622 Dr Reji Mathew Electrical Engineering UNSW Review of Motion Modelling and Estimation Introduction to Motion Modelling & Estimation Forward Motion Backward Motion Block Motion Estimation Motion

More information

Programmable Graphics Hardware (GPU) A Primer

Programmable Graphics Hardware (GPU) A Primer Programmable Graphics Hardware (GPU) A Primer Klaus Mueller Stony Brook University Computer Science Department Parallel Computing Explained video Parallel Computing Explained Any questions? Parallelism

More information

Implementation of Deep Convolutional Neural Net on a Digital Signal Processor

Implementation of Deep Convolutional Neural Net on a Digital Signal Processor Implementation of Deep Convolutional Neural Net on a Digital Signal Processor Elaina Chai December 12, 2014 1. Abstract In this paper I will discuss the feasibility of an implementation of an algorithm

More information

G P G P U : H I G H - P E R F O R M A N C E C O M P U T I N G

G P G P U : H I G H - P E R F O R M A N C E C O M P U T I N G Joined Advanced Student School (JASS) 2009 March 29 - April 7, 2009 St. Petersburg, Russia G P G P U : H I G H - P E R F O R M A N C E C O M P U T I N G Dmitry Puzyrev St. Petersburg State University Faculty

More information