QR refactoring #27
QR refactoring #27
Conversation
cortner
left a comment
cortner
left a comment
There was a problem hiding this comment.
just some questions, nothing blocking I think...
src/BPDual.jl
Outdated
| function Base.show(io::IO, t::ASPTracer) | ||
| nsteps = length(t.iteration) | ||
| nvars = isempty(t.solution) ? 0 : length(t.solution[end].nzind) | ||
| println(io, "ASPTracer with $nvars active variables at final step.") |
There was a problem hiding this comment.
I would output it a bit differently, something like
ASPTracer(maxactive = ...)
or something like that.
src/BPDual.jl
Outdated
| cur_r_size +=1 | ||
| cur_r_size +=1 | ||
| if itn % refactor_freq == 0 && cur_r_size > 0 | ||
| F = qr!(S[:, 1:cur_r_size]) |
There was a problem hiding this comment.
is it really safe to do this in-place?
There was a problem hiding this comment.
guess it would be better to copy?
There was a problem hiding this comment.
if in-place is safe then in-place is better. I just wanted to be certain that you are certain about it. But it overwrite S; so if S is just a buffer that gets overwritten again before the next qr! then it's fine.
src/BPDual.jl
Outdated
| cur_r_size -=1 | ||
| cur_r_size -=1 | ||
| if itn % refactor_freq == 0 && cur_r_size > 0 | ||
| F = qr!(S[:, 1:cur_r_size]) |
src/omp.jl
Outdated
| cur_r_size +=1 | ||
|
|
||
| if itn % refactor_freq == 0 && cur_r_size > 0 | ||
| F = qr!(S[:, 1:cur_r_size]) |
|
|
||
| # Solve the basis pursuit problem | ||
| tracer = asp_bpdn(A, b, 0.0, loglevel =0); | ||
| tracer = asp_bpdn(A, b, 0.0, loglevel =0, refactor_freq =20); |
There was a problem hiding this comment.
should there be tests with and without refactor? Could one design a test that requires a refactor?
There was a problem hiding this comment.
I guess we could add a test where the columns of A are nearly linearly dependent..
Added a refactor frequency parameter:
refactor_freq.TO DO: add the automatic triggering condition based on max(R_ii)/min(R_ii).