Doctorado-ML/stree
1
1
Fork 0
mirror of https://github.com/Doctorado-ML/STree.git synced 2025-08-15 23:46:02 +00:00
Code Issues Packages Projects Releases Wiki Activity
88 Commits 21 Branches 18 Tags
5cef0f48753dde283d717db6c01c4edd9343fa2a
Commit Graph

88 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Ricardo Montañana
5cef0f4875 Implement splitter type mutual info 2021-05-01 23:38:34 +02:00
Ricardo Montañana
28c7558f01 Update Readme 
Add max_features > n_features test
Add make doc
 2021-04-27 23:15:21 +02:00
Ricardo Montañana Gómez
e19d10f6a7 Package doc #7 (#34) 
* Add first doc info to sources

* Update doc to separate classes in api

* Refactor build_predictor

* Fix random_sate issue in non linear kernels

* Refactor score method using base class implementation

* Some quality refactoring

* Fix codecov config.

* Add sigmoid kernel

* Refactor setup and add Makefile
 2021-04-26 09:10:01 +02:00
Ricardo Montañana Gómez
02de394c96 Add select KBest features #17 (#35) 2021-04-26 01:48:50 +02:00
Ricardo Montañana Gómez
a4aac9d310 Create codeql-analysis.yml (#25) 2021-04-19 23:34:26 +02:00
Ricardo Montañana Gómez
8a18c998df Implement hyperparam. context based normalization (#32) 1.0 2021-04-18 18:57:39 +02:00
Ricardo Montañana
b55f59a3ec Fix compute number of nodes 2021-04-13 22:31:05 +02:00
Ricardo Montañana
783d105099 Add another nodes, leaves test 2021-04-09 10:56:54 +02:00
Ricardo Montañana
c36f685263 Fix unintended nested if in partition 2021-04-08 08:27:31 +02:00
Ricardo Montañana
0f89b044f1 Refactor train method 2021-04-07 01:02:30 +02:00
Ricardo Montañana Gómez
6ba973dfe1 Add a method that return nodes and leaves (#27) (#30) 
Add a test
Fix #27
 2021-03-23 14:30:32 +01:00
Ricardo Montañana Gómez
460c63a6d0 Fix depth sometimes is wrong (#26) (#29) 
Add a test to the tests set
Add depth to node description
Fix iterator and str test due to this addon
 2021-03-23 14:08:53 +01:00
Ricardo Montañana Gómez
f438124057 Fix mistakes (#24) (#28) 
Put pandas requirements in notebooks
clean requirements.txt
 2021-03-23 13:27:32 +01:00
Ricardo Montañana Gómez
147dad684c Weight0samples error (#23) 
* Add Hyperparameters description to README
Comment get_subspace method
Add environment info for binder (runtime.txt)

* Complete source comments
Change docstring type to numpy
update hyperameters table and explanation

* Fix problem with zero weighted samples
Solve WARNING: class label x specified in weight is not found
with a different approach

* Allow update of scikitlearn to latest version
1.0rc1 		2021-01-19 11:40:46 +01:00
Ricardo Montañana Gómez
3bdac9bd60 Complete source comments (#22) 
* Add Hyperparameters description to README
Comment get_subspace method
Add environment info for binder (runtime.txt)

* Complete source comments
Change docstring type to numpy
update hyperameters table and explanation

* Update Jupyter notebooks
 2021-01-19 10:44:59 +01:00
Ricardo Montañana Gómez
e4ac5075e5 Add main workflow action (#20) 
* Add main workflow action

* lock scikit-learn version to 0.23.2

* exchange codeship badge with githubs
 2021-01-11 13:46:30 +01:00
Ricardo Montañana Gómez
36816074ff Combinatorial explosion (#19) 
* Remove itertools combinations from subspaces

* Generates 5 random subspaces at most
 2021-01-10 13:32:22 +01:00
Ricardo Montañana
475ad7e752 Fix mistakes in function comments 2020-11-11 19:14:36 +01:00
Ricardo Montañana Gómez
1c869e154e Enhance partition (#16) 
#15 Create impurity function in Stree (consistent name, same criteria as other splitter parameter)
Create test for the new function
Update init test
Update test splitter parameters
Rename old impurity function to partition_impurity
close #15
* Complete implementation of splitter_type = impurity with tests
Remove max_distance & min_distance splitter types

* Fix mistake in computing multiclass node belief
Set default criterion for split to entropy instead of gini
Set default max_iter to 1e5 instead of 1e3
change up-down criterion to match SVC multiclass
Fix impurity method of splitting nodes
Update jupyter Notebooks
0.9rc6 		2020-11-03 11:36:05 +01:00
Ricardo Montañana
f5706c3159 Update version and notebooks 0.9rc5 2020-06-28 10:44:29 +02:00
Ricardo Montañana
be552fdd6c Add test for getting 3 feature_sets in Splitter 
Add ensemble notebook
 2020-06-28 02:45:08 +02:00
Ricardo Montañana
5e3a8e3ec5 Change adaboost notebook 2020-06-27 23:34:15 +02:00
Ricardo Montañana
554ec03c32 Get only 3 sets for best split 
Fix flaky test in Splitter_test
 2020-06-27 18:29:40 +02:00
Ricardo Montañana
4b7e4a3fb0 better solution to the sklearn bagging problem 
Add better tests
enhance .coveragerc
 2020-06-26 11:22:45 +02:00
Ricardo Montañana
76723993fd Solve Warning class label not found when bagging 2020-06-25 13:07:50 +02:00
Ricardo Montañana
ecd0b86f4d Solve the mistake of min and max distance 
The split criteria functions min and max distance return classes while
max_samples return distances positives and negatives to hyperplane of
the class with more samples in node
 2020-06-17 00:13:52 +02:00
Ricardo Montañana
3e52a4746c Fix entroy and information_gain functions 2020-06-16 13:56:02 +02:00
Ricardo Montañana Gómez
a20e45e8e7 Merge pull request #10 from Doctorado-ML/add_subspaces 
#2 Add subspaces
 2020-06-15 11:30:53 +02:00
Ricardo Montañana
9334951d1b #2 Cosmetic and style updates 2020-06-15 11:09:11 +02:00
Ricardo Montañana
736ab7ef20 #2 update benchmark notebook 2020-06-15 10:33:51 +02:00
Ricardo Montañana
c94bc068bd #2 Refactor Stree & create Splitter 
Add and test splitter parameter
 2020-06-15 00:22:57 +02:00
Ricardo Montañana
502ee72799 #2 Add predict and score support 
Add a test in features notebook
Show max_features in main.py
 2020-06-14 14:00:21 +02:00
Ricardo Montañana
f1ee4de37b #2 - Add gini and entropy measures 
rename get_dataset to load_dataset
add features and impurity to  __str__ of node
 2020-06-14 03:08:55 +02:00
Ricardo Montañana
ae1c199e21 # 2 - add max_features parameters 2020-06-13 17:58:45 +02:00
Ricardo Montañana
1bfe273a70 Fix problem in _min_distance 
Remove grapher (moved to another repo)
 2020-06-12 00:50:25 +02:00
Ricardo Montañana Gómez
647d21bdb5 Merge pull request #9 from Doctorado-ML/add_multiclass 
#6 Add multiclass
 2020-06-11 16:30:16 +02:00
Ricardo Montañana
1d392d534f #6 - Update tests and codecov conf 2020-06-11 13:45:24 +02:00
Ricardo Montañana
f360a2640c #6 - Add multiclass support 
Removed (by now) predict_proba. Created a notebook in jupyter
Added split_criteria parameter with min_distance and max_samples values
Refactor _distances
Refactor _split_criteria
Refactor _reorder_results
 2020-06-11 13:10:52 +02:00
Ricardo Montañana Gómez
45510b43bc Merge pull request #5 from Doctorado-ML/add_kernels 
#3 Add kernels to STree
 2020-06-09 13:43:31 +02:00
Ricardo Montañana
286a91a3d7 #3 refactor unneeded code and new test 2020-06-09 13:01:01 +02:00
Ricardo Montañana
5c31c2b2a5 #3 update features notebook 2020-06-09 02:12:56 +02:00
Ricardo Montañana
7e932de072 #3 Add sample_weights to score, update notebooks 
Update readme to use new names of notebooks
 2020-06-09 01:46:38 +02:00
Ricardo Montañana
26273e936a #3 Add degree hyperparam and update notebooks 
Update readme to add new  notebooks
 2020-06-08 20:16:42 +02:00
Ricardo Montañana
d7c0bc3bc5 #3 Complete multiclass in Stree 
Add multiclass dimensions management in distances method
Add gamma hyperparameter for non linear kernels
 2020-06-08 13:54:24 +02:00
Ricardo Montañana
3a48d8b405 #3 Rewrite some tests & remove use_predictions 
Remove use_predictions parameter as of now, the model always use it
 2020-06-08 01:51:21 +02:00
Ricardo Montañana
05b462716e #3 First try, change LinearSVC to SVC 
make a builder
start changing tests
 2020-06-07 20:26:59 +02:00
Ricardo Montañana
b824229121 #1 Add min_samples_split 
Fix #1
 2020-06-07 16:12:25 +02:00
Ricardo Montañana
8ba9b1b6a1 Remove travis ci and set codecov percentage 2020-06-06 19:47:00 +02:00
Ricardo Montañana
37577849db Fix parameter missing in method overload 2020-06-06 18:18:03 +02:00
Ricardo Montañana
cb10aea36e remove unneed test and cosmetic 2020-06-06 14:20:07 +02:00