. Let's spice things up with some Julia coding assignments, because clearly, solving dense matrices wasn't challenging enough 
.
https://venkovic.github.io/NLA-for-CS-and-IE.html #NumericalLinearAlgebra #JuliaCoding #Education #Humor #TechEducation #HackerNews #ngated
Numerical Linear Algebra Class in Julia TUM
One of many very good talks was given by Paolo Bientinesi from Umeå. He looked at how high level languages like Python, Julia or R make use of the excellent software ecosystem for #NumericalLinearAlgebra, given by LAPACK, BLAS and so on.
A simple test shows that they will execute the matrix product A*B*C always from the left and not choose order from the dimension (but Julia will fix this soon). Why is that bad?
I'm at a workshop in honour of zhaojun Bai, who was awarded an honorary doctorate from Stockholm University. Lot's of interesting talks.
However, can we all agree how awesome LAPACK is? It's quite old, but still state of the art!
I'm at #NumDiff17 in Halle this week and Daniel Kressner had this tidbit in his plenary on randomized linear algebra for differential equations:
The set of singular square matrices has Lebesgue measure zero, since it is characterized by a zero determinant. Thus, when you draw a random square matrix from, e.g., a normal distribution, it will be invertible with probability one. Neat!
At the Nordic Numerical Linear Algebra workshop, this year in Odense, Denmark. The second keynote by Alex Townsend from Cornell University via zoom. Worked very well, and was sustainable.
He says that you don't meet the Kahan matrix in a dark alley. Or otherwise put: Many worst case estimates are not representative of what happens in practice.
Also a very nice presentation of methods to compute the numerical rank of matrices.
At the Nordic Numerical Linear Algebra workshop, this year in Odense, Denmark. Interesting keynote by Erin Carson from Charles University about the state of the art on mixed precision iterative methods. Highly relevant in the new heterogeneous hardware world, e.g. the Hopper chip.
