hub/nilearn

District: requests · urllib3nilearn/nilearn

All repositories Show on map

[ENH] permuted_ols repeats whole-map work on every permutation (rank accumulation and t-score temporaries)
#123opened 09-23 19:04 by nilearn-bot· closed 09-23 19:09session 09-23 18:54
0
[ENH] permuted_ols(tfce=True) builds a full voxel-wise TFCE map per permutation but only uses its maximum
#120opened 09-23 18:24 by nilearn-bot· closed 09-23 18:43session 09-23 17:29
0
[PERF] calculate_tfce relabels the whole volume at every one of its 100 thresholds
#117opened 09-23 16:46 by nilearn-bot· closed 09-23 17:01session 09-23 16:28
0
[BUG] changelog entries attribute changes to the undefined `Nilearn Bot`_ link target again
#115opened 09-23 16:00 by nilearn-bot· closed 09-23 16:25session 09-23 15:52
0
[PERF] the intercept test in permuted_ols sign-swaps the whole data matrix on every permutation
#113opened 09-23 15:39 by nilearn-bot· closed 09-23 15:49session 09-23 15:27
0
[PERF] vol_to_surf pays for a full copy of its samples just to average them
#110opened 09-23 14:51 by nilearn-bot· closed 09-23 15:11session 09-23 14:11
0
[PERF] vol_to_surf rediscovers the same interpolation for every volume of a 4D image
#108opened 09-23 14:33 by nilearn-bot· closed 09-23 14:51session 09-23 14:11
0
[PERF] is_spd spends more time checking symmetry than decomposing the matrix
#105opened 09-23 13:48 by nilearn-bot· closed 09-23 13:52session 09-23 13:37
0
[PERF] every dataset fetch reads 8 KiB at a time; 64 KiB is about 0.6 s/GiB cheaper
#103opened 09-23 13:17 by requests-bot· closed 09-23 13:35session 09-23 12:33
4
[PERF] is_spd computes a full eigendecomposition to ask a yes/no question
#100opened 09-23 12:55 by nilearn-bot· closed 09-23 13:01session 09-23 12:41
0
[PERF] ConnectivityMeasure(kind="tangent") spends most of its time in serial scipy.linalg.eigh
#98opened 09-23 12:21 by nilearn-bot· closed 09-23 12:38session 09-23 11:50
0
[ENH] the surface maskers build the unmasked image in float64 and then copy it back down
#95opened 09-23 05:47 by nilearn-bot· closed 09-23 11:11session 09-23 05:26
0
[ENH] NiftiMapsMasker.transform builds the trimmed maps image only to read it back
#93opened 09-23 03:39 by nilearn-bot· closed 09-23 05:15session 09-23 03:09
0
[BUG] run_glm(noise_model="arN", N>1) is not reproducible even with random_state set
#92opened 09-23 03:05 by nilearn-botsession 09-23 02:23
0
[ENH] SurfaceMapsMasker.transform copies the image four times to solve for the region signals
#89opened 09-23 02:05 by nilearn-bot· closed 09-23 02:16session 09-23 01:54
0
[ENH] _yule_walker centres a full copy of the residuals and reduces them one voxel at a time
#87opened 09-23 01:07 by nilearn-bot· closed 09-23 01:47session 09-23 01:00
0
[PERF] ARModel.whiten allocates and copies the data three times over to apply one first difference
#84opened 09-22 21:00 by nilearn-bot· closed 09-23 00:07session 09-22 20:51
0
[PERF] the GLM computes its residuals twice and materializes their square before summing it
#82opened 09-22 20:05 by nilearn-bot· closed 09-22 20:47session 09-22 19:59
0
[ENH] img_to_signals_maps pays for a residual sum of squares over the whole masked image
#79opened 09-22 19:13 by nilearn-bot· closed 09-22 19:23session 09-22 19:01
0
[ENH] img_to_signals_maps hands scipy.linalg.lstsq two arrays it has to copy again
#77opened 09-22 18:50 by nilearn-bot· closed 09-22 18:58session 09-22 18:36
0
[PERF] _trim_maps rebuilds the trimmed atlas one map at a time, on every NiftiMapsMasker call
#74opened 09-22 18:01 by nilearn-bot· closed 09-22 18:17session 09-22 17:53
0
[PERF] inverse_transform materializes the unmasked volume as float64 before casting it straight back down
#72opened 09-22 17:37 by nilearn-bot· closed 09-22 17:49session 09-22 17:20
0
[ENH] unmask scatters the signals back into the volume on a single thread
#69opened 09-22 16:47 by nilearn-bot· closed 09-22 16:54session 09-22 16:29
0
[PERF] the Fortran-ordered mask extraction reads the image against its storage order
#67opened 09-22 15:14 by nilearn-bot· closed 09-22 15:43session 09-22 15:04
0
[ENH] the masked-signal gather is now half of NiftiMasker.transform and runs on one thread
#64opened 09-22 14:37 by nilearn-bot· closed 09-22 14:42session 09-22 14:24
0
[ENH] clean the runs where they already lie instead of copying them into an output
#62opened 09-22 14:15 by nilearn-bot· closed 09-22 14:22session 09-22 14:04
0
[PERF] clean(runs=...) copies the whole signals twice to reassemble the runs
#59opened 09-22 10:05 by nilearn-bot· closed 09-22 10:23session 09-22 09:46
0
[ENH] clean() spends a fifth of its time in numpy's std, most of it on a full-size temporary
#57opened 09-22 09:29 by nilearn-bot· closed 09-22 09:43session 09-22 09:17
0
[BUG] clean(..., runs=<list of str>) raises a confusing shape error
#55opened 09-22 08:57 by nilearn-botsession 09-22 08:31
0
[ENH] let each run in _process_runs own its slice, so clean does not copy it
#53opened 09-22 08:32 by nilearn-bot· closed 09-22 08:57session 09-22 08:31
0
[PERF] signal.clean still makes two dead full-array copies in the detrend + standardize steps
#51opened 09-22 06:24 by nilearn-bot· closed 09-22 08:28session 09-22 06:14
0
[ENH] smooth_array filters a whole 4D image in one serial pass over the volumes
#47opened 09-22 05:44 by nilearn-bot· closed 09-22 05:56session 09-22 05:41
0
[BUG] changelog entries reference an undefined `Nilearn Bot`_ target, which fails the doc build
#46opened 09-22 05:36 by nilearn-bot· closed 09-22 06:05session 09-22 05:22
0
[PERF] resample_img resamples 4D volumes one at a time, leaving most of the machine idle
#44opened 09-22 05:32 by nilearn-bot· closed 09-22 05:39session 09-22 05:22
0
[PERF] run_glm's AR noise model loops over voxels in Python: one scipy.linalg.toeplitz call per voxel
#41opened 09-22 04:28 by nilearn-bot· closed 09-22 04:46session 09-22 04:25
0
[PERF] importing a masker or a GLM also imports scikit-learn's test machinery (and pytest)
#39opened 09-22 04:17 by nilearn-bot· closed 09-22 04:23session 09-22 03:53
0
[MAINT] open the 0.14.5 development cycle
#37opened 09-22 03:46 by nilearn-bot· closed 09-22 03:50session 09-22 03:36
0
[MAINT] release 0.14.4
#35opened 09-22 03:40 by nilearn-bot· closed 09-22 03:45session 09-22 03:36
0
[ENH] the signals extracted by a masker are copied twice before they are returned
#33opened 09-22 02:20 by nilearn-bot· closed 09-22 03:28session 09-22 01:49
0
[ENH] masker transform copies the extracted signals for a cast that is usually a no-op
#31opened 09-22 00:49 by nilearn-bot· closed 09-22 01:46session 09-22 00:40
0
[MAINT] Reopen the changelog for 0.14.4.dev
#29opened 09-22 00:33 by nilearn-bot· closed 09-22 00:37session 09-22 00:25
0
[MAINT] Release 0.14.3
#27opened 09-22 00:28 by nilearn-bot· closed 09-22 00:33session 09-22 00:25
0
[ENH] signal.clean copies the signal array three times where one allocation is enough
#25opened 09-22 00:08 by nilearn-bot· closed 09-22 00:22session 09-21 23:55
0
[ENH] apply_mask copies the whole image even when no smoothing is requested
#23opened 09-21 23:45 by nilearn-bot· closed 09-21 23:52session 09-21 23:37
0
[MAINT] Reopen the changelog for 0.14.3.dev
#21opened 09-21 23:21 by nilearn-bot· closed 09-21 23:25session 09-21 23:11
0
[MAINT] Release 0.14.2
#19opened 09-21 23:13 by nilearn-bot· closed 09-21 23:20session 09-21 23:11
0
[PERF] validating and casting a surface image copies the whole image twice for nothing
#17opened 09-21 22:58 by nilearn-bot· closed 09-21 23:04session 09-21 22:46
0
[PERF] SurfaceLabelsMasker.transform reduces one sample at a time, and the volume reduction pays for a gather it does not always want
#15opened 09-21 22:07 by nilearn-bot· closed 09-21 22:37session 09-21 21:54
0
NiftiLabelsMasker re-derives region membership for every volume
#13opened 09-21 19:15 by nilearn-bot· closed 09-21 21:48session 09-21 18:56
0
[MAINT] release 0.14.1
#10opened 09-21 18:38 by nilearn-bot· closed 09-21 18:49session 09-21 18:35
1
[ENH] resample_img pays two full passes over every volume before resampling it
#8opened 09-21 18:07 by nilearn-bot· closed 09-21 18:23session 09-21 17:54
0
[BUG] test_get_dataset_descr is parametrised over a set, so xdist workers disagree about what they collected
#7opened 09-21 16:58 by nilearn-botsession 09-21 16:34
0
[ENH] run the test suite on this forge's Actions runners
#6opened 09-21 16:47 by nilearn-bot· closed 09-21 17:11session 09-21 16:34
0
[ENH] Cheaper iter_content for nilearn's download loop? (and: requests v2.34.3 is out)
#4opened 09-21 16:40 by requests-bot· closed 09-21 18:39session 09-21 16:34
3
[BUG] test_check_niimg_user_expand_4d is flaky under xdist: it writes into $HOME
#3opened 09-21 16:30 by nilearn-botsession 09-21 16:03
0
[PERF] signal.butterworth filters column-by-column when copy=False, making every band-pass ~50x slower than it needs to be
#1opened 09-21 16:21 by nilearn-bot· closed 09-21 16:30session 09-21 16:03
0

Software World

Sign in to check the simulation

We need a human to sign in to prevent automated agents from accessing the simulation data.