Sébastien Loisel

Constrained optimization

5

How to implement constrained optimization? - Which algorithm to use? Barrier method? - How much derivative information is needed for constraint function(s)? How to pass constraint function(s)? Is it an...

Eigenvalues and singular values of complex matrices

1

Need to have implementations of eig and svd for complex matrices.

More example applications

4

Example applications: I have small toy demos in the workshop but if there are larger applications outside the workshop, it would be nice to link to them.

Test numeric.solveLP()

2

Test numeric.solveLP() against standard test problems. Right now, it's only tested against randomly generated L1 norm minimization problems.

Typed array matrices

9

How is a typed array matrix specified? Is it an Array of typed Arrays? Or is it a single flat typed array? Is it reasonable to take a flat Float64Array...

Here it is: http://numericjs.com/workshop.php?link=95813bd4244ff7247357b2035591da70c88a0e80086a6e3f1d53a9de58295f53

Bug in numeric.eig()

1

numeric.eig() on the identity returns funny results. This is because the handling of multiple eigenvalues is sort of weird right now. This is a bug and it should be fixed.

Automatic differentiation

2

Should this be implemented using an AD object? Can this AD object handle linear algebra? How much of it is orthogonal to numeric?

Implement CG. http://en.wikipedia.org/wiki/Conjugate_gradient_method The implementation should be matrix-free and should allow for a matrix-free preconditioner.

Implement GMRES. http://en.wikipedia.org/wiki/Generalized_minimal_residual_method The implementation should be matrix-free and allow for a matrix-free preconditioner.