feat: Add getters for naive bayes structs #74

morenol · 2021-01-22T21:07:47Z

This is related to #73:

Things that are done in this PR

Add a getter equivalent to the scikitlearn attributes of the GaussianNB class for classes_, class_prior_, var_, and theta_.
Rename sigma to var, in order to have a similar naming that in scikitlearn (sigma is deprecated)

codecov-io · 2021-01-22T23:55:15Z

Codecov Report

Merging #74 (aa4c144) into development (bd5fbb6) will increase coverage by 0.05%.
The diff coverage is 93.05%.

@@               Coverage Diff               @@
##           development      #74      +/-   ##
===============================================
+ Coverage        83.77%   83.83%   +0.05%     
===============================================
  Files               75       75              
  Lines             7803     7850      +47     
===============================================
+ Hits              6537     6581      +44     
- Misses            1266     1269       +3

Impacted Files	Coverage Δ
src/naive_bayes/categorical.rs	`76.51% <66.66%> (-0.23%)`	⬇️
src/naive_bayes/bernoulli.rs	`71.42% <92.30%> (+1.42%)`	⬆️
src/naive_bayes/gaussian.rs	`85.18% <93.10%> (+0.97%)`	⬆️
src/naive_bayes/multinomial.rs	`75.63% <96.29%> (+4.49%)`	⬆️
src/svm/svc.rs	`89.93% <0.00%> (+0.31%)`	⬆️

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 bd5fbb6...aa4c144. Read the comment docs.

VolodymyrOrlov · 2021-01-26T17:59:45Z

Hi @morenol. Is this PR still a work in progress or I can review it?

morenol · 2021-01-26T21:08:00Z

Hi @morenol. Is this PR still a work in progress or I can review it?

Hi, @VolodymyrOrlov this is still a work in progress. I'll let you know when this is ready to review

morenol · 2021-01-31T20:15:29Z

src/naive_bayes/gaussian.rs

@@ -223,6 +219,30 @@ impl<T: RealNumber, M: Matrix<T>> GaussianNB<T, M> {
    pub fn predict(&self, x: &M) -> Result<M::RowVector, Failed> {
        self.inner.predict(x)
    }
+


Missing:

epsilon_

https://github.com/scikit-learn/scikit-learn/blob/88be2abb7b7f7450dc569e0065e672a0676e4130/sklearn/naive_bayes.py#L143

morenol · 2021-01-31T20:18:17Z

src/naive_bayes/bernoulli.rs

@@ -266,6 +266,12 @@ impl<T: RealNumber, M: Matrix<T>> BernoulliNB<T, M> {
            self.inner.predict(x)
        }
    }
+
+    /// Class labels known to the classifier.


Missing:
- [ ] class_log_prior_
- [x] feature_log_prob_

https://github.com/scikit-learn/scikit-learn/blob/88be2abb7b7f7450dc569e0065e672a0676e4130/sklearn/naive_bayes.py#L970

morenol · 2021-01-31T20:20:28Z

src/naive_bayes/multinomial.rs

@@ -240,6 +243,18 @@ impl<T: RealNumber, M: Matrix<T>> MultinomialNB<T, M> {
    pub fn predict(&self, x: &M) -> Result<M::RowVector, Failed> {
        self.inner.predict(x)
    }
+


Missing:
- [ ] class_log_prior_

https://github.com/scikit-learn/scikit-learn/blob/88be2abb7b7f7450dc569e0065e672a0676e4130/sklearn/naive_bayes.py#L702

morenol · 2021-01-31T20:23:23Z

src/naive_bayes/categorical.rs

+    /// Class labels known to the classifier.
+    /// Returns a vector of size n_classes.
+    pub fn classes(&self) -> &Vec<T> {
+        &self.inner.distribution.class_labels


https://github.com/scikit-learn/scikit-learn/blob/88be2abb7b7f7450dc569e0065e672a0676e4130/sklearn/naive_bayes.py#L1112

category_count_

class_log_prior_

feature_log_prob_

morenol · 2021-02-15T02:39:05Z

@VolodymyrOrlov Could you kindly review this now? I implemented most of the getters from scikit where we don't need to do extra computations and just access internal data from the structs. For instance, Right now we are storing the class_priors as a probability and in order to get class_log_prior_ we will need to compute the ln of each probability, should we compute that on runtime or store them too in the internal structure? or just ignore those getters for now?

VolodymyrOrlov

lgtm!

VolodymyrOrlov · 2021-02-16T17:56:52Z

@VolodymyrOrlov Could you kindly review this now? I implemented most of the getters from scikit where we don't need to do extra computations and just access internal data from the structs. For instance, Right now we are storing the class_priors as a probability and in order to get class_log_prior_ we will need to compute the ln of each probability, should we compute that on runtime or store them too in the internal structure? or just ignore those getters for now?

Some users might find class_log_prior_ useful but most will likely not need it. We don't have to be 100% compatible with Scikit learn, especially when it comes to a minor convenience methods like the one that returns class probabilities. Besides, we can always add it later if someone requests it.

Add classes getter to CategoricalNB Add classes getter to MultinomialNB

morenol force-pushed the lmm/naive_bayes_getters branch from 823dee3 to 64548dc Compare January 22, 2021 23:50

morenol force-pushed the lmm/naive_bayes_getters branch 2 times, most recently from 30086a5 to 06fda75 Compare January 23, 2021 19:39

morenol commented Jan 31, 2021

View reviewed changes

morenol force-pushed the lmm/naive_bayes_getters branch 4 times, most recently from b81111d to ed1511c Compare February 1, 2021 00:18

morenol force-pushed the lmm/naive_bayes_getters branch from ed1511c to ecf2b44 Compare February 15, 2021 01:32

morenol force-pushed the lmm/naive_bayes_getters branch from 4d0fb24 to ae3e3ee Compare February 15, 2021 02:39

morenol changed the title ~~[WIP] feat: Add getters for naive bayes structs~~ feat: Add getters for naive bayes structs Feb 15, 2021

morenol requested a review from VolodymyrOrlov February 15, 2021 02:39

VolodymyrOrlov previously approved these changes Feb 16, 2021

View reviewed changes

morenol added 9 commits February 16, 2021 18:20

feat: Add getters for GaussianNB

98c2343

Add classes getter to BernoulliNB

7e66a56

Add classes getter to CategoricalNB Add classes getter to MultinomialNB

Add feature_log_prob getter to MultinomialNB

8ec2f8d

Add class_count to NB structs

24f4b00

Add n_features getter for NB

bfd1e01

Add feature_count to MultinomialNB and BernoulliNB

182086d

Add n_categories to CategoricalNB

0923c86

Implement feature_log_prob and category_count getter for CategoricalNB

4f95ac4

Implement feature_log_prob for BernoulliNB

2374873

Return Failed error instead of panicking

04aad18

morenol dismissed VolodymyrOrlov’s stale review via 04aad18 February 16, 2021 22:20

morenol force-pushed the lmm/naive_bayes_getters branch from f1a55d0 to 04aad18 Compare February 16, 2021 22:20

Fix clippy warning

5d3c31f

morenol requested a review from VolodymyrOrlov February 16, 2021 22:36

VolodymyrOrlov approved these changes Feb 25, 2021

View reviewed changes

morenol merged commit 1b42f8a into smartcorelib:development Feb 25, 2021

morenol deleted the lmm/naive_bayes_getters branch February 25, 2021 19:45

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: Add getters for naive bayes structs #74

feat: Add getters for naive bayes structs #74

Uh oh!

morenol commented Jan 22, 2021

Uh oh!

codecov-io commented Jan 22, 2021 •

edited

Loading

Uh oh!

VolodymyrOrlov commented Jan 26, 2021

Uh oh!

morenol commented Jan 26, 2021 •

edited

Loading

Uh oh!

morenol Jan 31, 2021 •

edited

Loading

Uh oh!

morenol Jan 31, 2021 •

edited

Loading

Uh oh!

morenol Jan 31, 2021 •

edited

Loading

Uh oh!

morenol Jan 31, 2021 •

edited

Loading

Uh oh!

morenol commented Feb 15, 2021 •

edited

Loading

Uh oh!

VolodymyrOrlov left a comment

Uh oh!

VolodymyrOrlov commented Feb 16, 2021

Uh oh!

Uh oh!

feat: Add getters for naive bayes structs #74

feat: Add getters for naive bayes structs #74

Uh oh!

Conversation

morenol commented Jan 22, 2021

Uh oh!

codecov-io commented Jan 22, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

VolodymyrOrlov commented Jan 26, 2021

Uh oh!

morenol commented Jan 26, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

morenol Jan 31, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

morenol Jan 31, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

morenol Jan 31, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

morenol Jan 31, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

morenol commented Feb 15, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

VolodymyrOrlov left a comment

Choose a reason for hiding this comment

Uh oh!

VolodymyrOrlov commented Feb 16, 2021

Uh oh!

Uh oh!

codecov-io commented Jan 22, 2021 •

edited

Loading

morenol commented Jan 26, 2021 •

edited

Loading

morenol Jan 31, 2021 •

edited

Loading

morenol Jan 31, 2021 •

edited

Loading

morenol Jan 31, 2021 •

edited

Loading

morenol Jan 31, 2021 •

edited

Loading

morenol commented Feb 15, 2021 •

edited

Loading