issues
search
djsutherland
/
py-sdm
Python implementation of nonparametric nearest-neighbor-based estimators for divergences between distributions.
http://cs.cmu.edu/~dsutherl/sdm/
BSD 3-Clause "New" or "Revised" License
48
stars
8
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
be better about random states
#53
djsutherland
closed
9 years ago
0
Bug in _load_features
#52
cjauvin
closed
9 years ago
1
streaming support
#51
djsutherland
closed
10 years ago
1
results not completely reproducible
#50
djsutherland
opened
10 years ago
0
make it work for one-dimensional features
#49
djsutherland
opened
10 years ago
0
don't require numpy/scipy to do setup.py egg_info, etc
#48
djsutherland
opened
10 years ago
0
python 3 support
#47
djsutherland
opened
10 years ago
0
improve sdm CLI format handling
#46
djsutherland
opened
10 years ago
0
better divergence cache
#45
djsutherland
opened
10 years ago
0
flann linking issues
#44
djsutherland
closed
10 years ago
1
add nyström support
#43
djsutherland
opened
10 years ago
0
distributed divergence support
#42
djsutherland
opened
10 years ago
0
support other, non-SDM-style set classifiers
#41
djsutherland
opened
10 years ago
0
performance on scene dataset sucks
#40
djsutherland
opened
10 years ago
0
Features class: don't require it to be stacked
#39
djsutherland
opened
10 years ago
0
support python-format in "svm predict" commandline
#38
djsutherland
closed
10 years ago
1
hang in div calculation sometimes
#37
djsutherland
opened
10 years ago
0
SDM classes should save indices, rhos for training data to speed up repeated test evaluation
#36
djsutherland
opened
10 years ago
0
transduction doesn't work when labels array is boolean
#35
djsutherland
opened
10 years ago
0
friendlier error messages / format checking if features file format is wrong
#34
djsutherland
opened
10 years ago
0
get flann location from cyflann; possibly also support FLANN_DIR
#33
djsutherland
closed
11 years ago
0
build failure
#32
wgmueller1
closed
11 years ago
1
sparse format for divergences
#31
djsutherland
opened
11 years ago
1
multiprocessing seems to sometimes flake out
#30
djsutherland
opened
11 years ago
1
cache the projected matrices
#29
djsutherland
opened
11 years ago
0
Features.load_from_perbag not loading extras properly, apparently
#28
djsutherland
closed
11 years ago
1
"crossvalidate" for one-class
#27
djsutherland
opened
11 years ago
0
np_divs: fix progressbar
#26
djsutherland
closed
11 years ago
0
np_divs: only make indices / get rhos if we need them
#25
djsutherland
opened
11 years ago
0
handle default feature names better
#24
djsutherland
opened
11 years ago
1
more thorough tests
#23
djsutherland
opened
11 years ago
1
more robust divergence caching
#22
djsutherland
opened
11 years ago
1
optionally do tuning inductively
#21
djsutherland
opened
11 years ago
0
fill in output as we go in estimate_divs, maybe elsewhere
#20
djsutherland
opened
11 years ago
0
use flann's multithreading instead of multiprocessing in get_divs
#19
djsutherland
closed
11 years ago
1
support different formats for raw features
#18
djsutherland
closed
11 years ago
0
more flexibility in choosing images for extract_features
#17
djsutherland
closed
11 years ago
0
support one-class SVMs
#16
djsutherland
closed
11 years ago
1
SDMs should support decision values / probabilities
#15
djsutherland
closed
11 years ago
1
`crossvalidate` should optionally return more info about the folds
#14
djsutherland
closed
11 years ago
1
allow specifying exactly what CV folds to use
#13
djsutherland
closed
11 years ago
1
support CVing through k, multiple div funcs (values of alpha, ...)
#12
djsutherland
opened
11 years ago
0
smarter projection when we're doing it inductively
#11
djsutherland
opened
11 years ago
2
randomize output of parameter tuning when there are ties
#10
djsutherland
closed
11 years ago
0
support non-Gaussian kernels
#9
djsutherland
opened
11 years ago
1
try only getting negative eigenvectors for projection
#8
djsutherland
opened
11 years ago
0
allow repeated CVs in same sdm.py call
#7
djsutherland
opened
11 years ago
0
proc_features should save PCA transformation
#6
djsutherland
closed
11 years ago
0
get rid of "labels" key in the div file format, switch to label/name
#5
djsutherland
closed
11 years ago
0
check that image featurization is correct
#4
djsutherland
opened
11 years ago
0
Next