multiple updates

[yangarbiter]

Added query strategies

• density weighted meta algorithm
• CSRPE
• add some cost functions in libact.utils.multilabel

Interface

• _get_scores for query_strategy
• clone method for Model

Bug fix

• density_weighted_uncertainty to documentation

Unit test improvements

• scikit-learn after 0.20, the output of logistic_regression.predict_proba changes a little which makes some of the test to fail, we fixed scikit-learn version to 0.19.2 for now
• fix the solver and multi_class attribute for LogisticRegression and gamma for SVC
• test for multi-label criteria

Other improvements

• Update the installation process, added the options to not compiling C-extension, which seems to make a lot of windows user fail the installation process.

[codecov-io]

Codecov Report

Merging #153 into master will decrease coverage by 0.22%.
The diff coverage is 88.62%.

@@            Coverage Diff             @@
##           master     #153      +/-   ##
==========================================
- Coverage   89.69%   89.46%   -0.23%     
==========================================
  Files          34       37       +3     
  Lines        1358     1557     +199     
==========================================
+ Hits         1218     1393     +175     
- Misses        140      164      +24

Impacted Files	Coverage Δ
...query_strategies/multilabel/binary_minimization.py	100% <ø> (ø)	⬆️
...es/multilabel/multilabel_with_auxiliary_learner.py	89.58% <ø> (-0.22%)	⬇️
..._strategies/multilabel/adaptive_active_learning.py	100% <ø> (ø)	⬆️
..._strategies/multilabel/maximum_margin_reduction.py	97.77% <ø> (ø)	⬆️
..._strategies/multiclass/expected_error_reduction.py	92.5% <ø> (ø)	⬆️
...ery_strategies/multiclass/hierarchical_sampling.py	95.85% <100%> (ø)	⬆️
libact/query_strategies/multilabel/__init__.py	100% <100%> (ø)	⬆️
libact/utils/multilabel/__init__.py	100% <100%> (ø)
libact/query_strategies/uncertainty_sampling.py	95.23% <100%> (+0.64%)	⬆️
libact/base/interfaces.py	73.52% <50%> (-1.48%)	⬇️
... and 8 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e37e9ed...6b380e2. Read the comment docs.