CMU 15-418 (Spring 2012) Final Project:

Principal Component Analysis on the GPU

Maxwell Koo (mjkoo)

Reports

Working Schedule

Week | What We Plan To Do | What We Actually Did |

Apr 8-14 | Research PCA algorithms and any existing GPU implementations | Worked on sample facial recognition app, using linear algebra code from numpy and OpenCV |

Apr 15-21 | Begin working on a GPU-based simulation, comparing to the results from the CPU implementation | Examined CUDA implementations of linear algebra code, evaulated which operations are parallelizable |

Apr 22-28 | Optimization | Worked on optimizing the matrix normalization, transpose, and multiplication steps before Eigenvector decomposition takes place |

Apr 29-May 5 | Optimization | Worked on optmizing for memory accesses using concepts of spacial locality, combine transpose and multiplication steps |

May 6-11 | Create a report with my findings | Created a report with my findings |

Working Log

4/10/12 - Switched to new project idea, began background research