gibson:teaching:fall-2014:iam961

Instructor: John Gibson, john.gibson@unh.edu

Office hours: Tue 2-3pm, Thu 12-1pm, Kingsbury N309E, or after class

Numerical linear algebra is the science of solving systems of linear equations and the eigenvalue problem on a digital computer –problems are at the root of the vast bulk of scientific computation. Compared to classical linear algebra, the finite precision and computational cost of numerical mathematics brings in a number of important new concepts, including conditioning, stability, and accuracy, and efficiency. We will develop these ideas and learn the most important numerical linear algebra algorithms: QR, LU, SVD decompositions, Gramm-Schmidt orthogonalization, the QR eigenvalue algorithm, and Krylov subspace methods. Time permitting, we will also study key algorithms for function optimization and the solution of systems of nonlinear equations.

Text: *Numerical Linear Algebra*, by Trefethen and Bau, SIAM Press.

lecture notes | date |
---|---|

SVD demo | 09/24 |

HW | due | topic | comments |
---|---|---|---|

HW1 | 9/22 | norms | |

HW2 | 10/06 | SVD | |

HW3 | 10/20 | QR algorithms | |

HW4 | 11/17 | conditioning, stability, and accuracy | |

HW5 | 12/15 | Rayleigh quotient iteration and Krylov subspace methods |

exam | date | comments |
---|---|---|

final | t.b.d. |

gibson/teaching/fall-2014/iam961.txt · Last modified: 2014/12/05 13:47 by gibson