Curtis G. Northcutt
|
3984417c01
|
Update for GNU GPL licensing
|
2021-03-03 18:08:34 -05:00 |
|
cgnorthcutt
|
e647398bd2
|
BIG UPDATE: fully support Windows multiprocessing and python 3.7
|
2020-02-17 01:59:33 -05:00 |
|
cgnorthcutt
|
b1226fe4da
|
PEP-8 adherence
|
2020-02-16 21:19:41 -05:00 |
|
cgnorthcutt
|
f2f7a3ce29
|
Make Windows Python pruning default to 1 process and not use multiprocessing
|
2020-02-14 15:56:52 -05:00 |
|
cgnorthcutt
|
a6b28030da
|
simplify logic and shorten code for prune_by_noise_rate
|
2019-09-19 12:22:44 -04:00 |
|
cgnorthcutt
|
5ec1a2cf81
|
FATAL bug fix for prune_by_noise_rate. margin was wrong. was same as np.argmax
|
2019-09-19 12:00:01 -04:00 |
|
Curtis Northcutt
|
1aed258481
|
Remove unused parameter
|
2019-09-04 21:55:58 -04:00 |
|
Curtis Northcutt
|
b7af0bba86
|
BUG FIX: wrong index was used for s_counts check
|
2019-09-04 16:52:47 -04:00 |
|
Curtis Northcutt
|
54022a7e98
|
Fixed bug: make data global for _prune_by methods
|
2019-09-04 00:08:22 -04:00 |
|
Curtis Northcutt
|
4a1e7cd195
|
Support non-parallelized version of pruning with n_jobs = 1
|
2019-09-03 23:28:40 -04:00 |
|
Curtis G. Northcutt
|
1625d9e9dd
|
Major speed-up by removing np.all on single labels
|
2019-08-21 01:07:51 -04:00 |
|
cgnorthcutt
|
4fc013aaec
|
remove ipynb [ ] artifacts
|
2019-08-19 18:02:42 -04:00 |
|
Curtis Northcutt
|
6d410511f8
|
Added verbose parameter, default printing off
|
2019-07-27 21:42:58 -07:00 |
|
Curtis Northcutt
|
8892a99cb0
|
Remove a multiprocessing method from coverage
|
2019-07-27 18:54:17 -07:00 |
|
Curtis Northcutt
|
361a7dbb40
|
Remove multiprocessing methods from coverage
|
2019-07-27 18:45:00 -07:00 |
|
Curtis Northcutt
|
976bde7a1c
|
BIG CHANGE: added multi_label support, fixed cj rounding, added tests
|
2019-07-27 18:23:11 -07:00 |
|
Curtis Northcutt
|
1b04c7c346
|
remove calls to estimate_confident_joint_from_probabilities
|
2019-07-26 22:23:36 -04:00 |
|
cgnorthcutt
|
f19c6b5792
|
Fixed warning message.
|
2019-07-02 15:05:18 -04:00 |
|
Curtis Northcutt
|
07833d133f
|
Fix order of imports for pylint compliance
|
2019-06-24 16:43:39 -04:00 |
|
Curtis Northcutt
|
82596251d0
|
Made pruning code flake8 pylint compliant.
|
2019-06-24 16:06:39 -04:00 |
|
Curtis Northcutt
|
7a12ecc938
|
Fixed test. only use exact pruning count with prune_by_class
|
2019-06-24 14:56:59 -04:00 |
|
Curtis Northcutt
|
878925ad8e
|
Fix python 2 and 3 compatibility with stdout.flush
|
2019-06-24 12:44:05 -04:00 |
|
Curtis Northcutt
|
96262c0c39
|
Error Fix: remove tqdm import
|
2019-06-19 17:36:04 -04:00 |
|
Curtis Northcutt
|
f33c97e720
|
BIG CHANGE: Remove tqdm as a package dependency. instead warn to install.
|
2019-06-19 17:15:37 -04:00 |
|
cgnorthcutt
|
95228b3e4d
|
MAJOR CHANGES: remove maxpred==label, multi-label support, simplified ordering
|
2019-02-26 16:55:04 -05:00 |
|
cgnorthcutt
|
1c7d341a9a
|
MAJOR CHANGE: multiprocessed ordering, removed prune_count_method
|
2019-02-16 03:51:24 -05:00 |
|
cgnorthcutt
|
f44372cf31
|
Return array of False (isntead of None) when no label errors
|
2019-02-15 22:44:56 -05:00 |
|
cgnorthcutt
|
e7c961a9cc
|
BIG CHANGE: Confident learning now uses multiprocessing for finding label errors.
|
2019-02-15 22:23:58 -05:00 |
|
cgnorthcutt
|
574346e7ab
|
Add environment bang header.
|
2019-02-14 19:27:16 -05:00 |
|
cgnorthcutt
|
66de6a42fc
|
Update docstsring to explain multi_label param
|
2019-02-14 18:35:04 -05:00 |
|
Curtis Northcutt
|
6658e483c5
|
Add support for multi-label classification.
|
2019-01-21 20:59:56 -05:00 |
|
Curtis Northcutt
|
49af85e6ae
|
Fixed error that occurs when num2prune == len(s)
|
2018-12-24 13:03:53 -08:00 |
|
Curtis Northcutt
|
f844042488
|
Fixed typo.
|
2018-12-22 23:19:09 -08:00 |
|
Curtis Northcutt
|
cac0dc2b06
|
Fixed typo.
|
2018-12-22 23:14:11 -08:00 |
|
Curtis Northcutt
|
de53acc7b4
|
Added feature to return sorted label error indices in get_noise_indices
|
2018-12-22 23:12:17 -08:00 |
|
Curtis Northcutt
|
92cb914a94
|
Fixed broadcasting error.
|
2018-12-22 22:31:40 -08:00 |
|
Curtis Northcutt
|
ed34e5895c
|
prune_by_noise_rate now orders by p(true) - p(noisy) instead of p(true)
|
2018-12-22 22:22:15 -08:00 |
|
Curtis Northcutt
|
1607be6f37
|
Optimization speed increase -- don't find errors if no errors to find
|
2018-12-22 22:08:31 -08:00 |
|
Curtis Northcutt
|
0df1869c0e
|
Fixed num_to_remove_per_class
|
2018-12-07 22:46:39 -05:00 |
|
Curtis Northcutt
|
88854d73d6
|
Explain which prune_count_method to use in comments.
|
2018-12-06 18:02:34 -05:00 |
|
Curtis Northcutt
|
5c16727068
|
Finds the indices of the examples with label errors in a dataset.
|
2018-11-21 12:52:09 -05:00 |
|