Activity for Mark Borgerding
-
Mark Borgerding posted a comment on discussion Help
Yes, it is to prevent overflow.
-
The fixed point FFT and IFFT both scale downwards to make overflow impossible.
-
btw .. you can't just "zero some coefficients in the FFT transform" without bad side...
-
Yes. It certainly is possible. A search for "fft 2d convolution" brings up several...
-
I don't have the time right now to dig into your code. My first suspicion is the...
-
I cannot think of any real limitations of KISSFFT in this regard, except perhaps...
-
BGZeroMeanEstimIn: fixed estimInit
-
Edit: This was a bug in vorbis-tools (oggdec). See my "On closer examination" post...
-
OK, on closer examination, it looks like this is a problem with vorbis-tools. Under...
-
Here's the extra version info as was earlier suggested. It fails consistently under...
-
I observe consistent (100%) failure on Centos7 and the two most recent Ubuntu versions...
-
I observe consistent (100%) failure on Centos7 and Ubuntu 14.04 when trying to pipe...
-
Here's the extra version info as was earlier suggested. Under CentOS Linux release...
-
I observe consistent (100%) failure on Centos7 and Ubuntu 14.10 when trying to pipe...
-
moved SNIPE-specific scripts into SNIPE-specifi...
-
comparison of SNIPE with "overmatched" Bernoull...
-
comparison of MultiSNIPEstim vs a matched prior...
-
moved test script to the test directory
-
version of SNIPE that allows multiple delta fun...
-
prevent infinite rvar (divide-by-zero) while st...
-
fixed negative KL divergence for Sum-Product ve...
-
fix for logScale with finite xvarBig
-
missing term in logScale can be calculated when...
-
SNIPEstim: added option for finite variance on ...
-
corrected and expanded comments regarding NullE...
-
explicitly project the answers onto the real ax...
-
I suspect you are not initializing all your input. Bob bob90-06@users.sf.net wrote:...
-
reworked testSNIPE.m. fixed NullEstimIn for co...
-
NullEstimIn: added maxSumVal and isCmplx flags.
-
SNIPEstim: implemented logScale for complex case
-
fixed xvar for complex case
-
fixed complex SNIPE case
-
SNIPE now provides a meaningful negative KL div...
-
corrected comment
-
updated logScale to be more complete and used t...
-
if all iterations pass, then the second trace i...
-
When determining whether a cast to MatrixLinTra...
-
Did you look at the derivation pdf available under "files"? I'd say it is pretty...
-
generalized SNIPE to allow the small values to ...
-
added a convenience method to LinTrans that cal...
-
when an impossible intersection occurs, clippin...
-
added Sum-Product (MMSE) option to UnifEstimIn
-
gampEstBasic now works with stopFcn and stopFcn...
-
GampOpt.warmStartCopy compatible with the estFi...
-
adapt to nargout=3 calling of estimInit by gamp...
-
added SNIPEstim methods to find omega that prov...
-
added stopFcn2: gives user a view of much more ...
-
GM_update better at handling complex input
-
added gampOpt.histIntvl=K to allow periodic sna...
-
added TruthReporter utility class to wrap an in...
-
added GetGroupIndices function to return a cell...
-
show magnitude of correlation coefficients for ...
-
show sign of correlation coefficients. Added co...
-
Added grids. Tweaked line styles to be more con...
-
Graphically show per-iteration historical data ...
-
That sounds really cool. I think Eigen eigen.tuxfamily.org would be a natural place...
-
OK to concatenate result of adjoint operator ' ...
-
Perhaps there is a scaling difference between kissfft and Audacity?
-
Are you sure you are using the float version of kissfft? cx_out should be complex,...
-
renamed SNIPE to SNIPEstim
-
renaming SNIPE.m to SNIPEstim.m
-
SNIPE estimator (shrinkage) function derived in...
-
added warmstart fields related to adaptive step...
-
added warmStartCopy method to GampOpt to make i...
1