New Feature: plotting

.gitignore

@@ -6,3 +6,4 @@ build/
 _glmnetmodule.c
 dist
 MANIFEST
+.ipynb_checkpoints

CHANGELOG.md

@@ -3,6 +3,10 @@ All notable changes to this project will be documented in this file.
 This project adheres to [Semantic Versioning](http://semver.org/).
 
 ## Unreleased
+## 2.2.2 - 2021-03-21
+### Added
+* [#70](https://github.com/civisanalytics/python-glmnet/pull/70)
+  Add new plotting functionality for cross-validation score and coefficient path.
 
 ## 2.2.1 - 2020-06-30
 ### Fix

dev-requirements.txt

@@ -3,3 +3,5 @@ numpy==1.19.0
 pytest==5.4.3
 scikit-learn==0.23.1
 scipy==1.5.0
+pandas==1.0.1
+matplotlib==3.3.4

glmnet/linear.py

@@ -14,7 +14,7 @@
                          _check_user_lambda,
                          _interpolate_model,
                          _score_lambda_path)
-
+from glmnet.plotting import coeff_path_plot, cv_score_plot
 
 class ElasticNet(BaseEstimator):
     """Elastic Net with squared error loss.
@@ -198,6 +198,8 @@ def fit(self, X, y, sample_weight=None, relative_penalties=None, groups=None):
         self : object
             Returns self.
         """
+        # keep local copy of features
+        self.X_ = X        
 
         X, y = check_X_y(X, y, accept_sparse='csr', ensure_min_samples=2)
         if sample_weight is None:
@@ -446,3 +448,150 @@ def r2_reverse(y_pred, y_true):
 
         # compute the score for each value of lambda
         return np.apply_along_axis(r2_reverse, 0, pred, y)
+
+    def plot_coeff_path(self, feature_names=None, figsize=(10, 6), linestyle="-", fontsize=18,
+                        grid=True, legend=True, legendloc="center", xlabel=None, ylabel=None, 
+                        title=None, yscale=None, bbox_to_anchor=None, save_path=None):
+        """Plot coefficient's paths vs -Log(lambda).
+
+        Parameters
+        ----------
+        feature_names : list, shape (n_features,)
+            Input features names neede for legend.
+
+        figsize : tuple or list, as (width, height)
+            Figure size.
+
+        linestyle: string
+            Linestyle of coefficients' paths.
+
+        fontsize : int, float
+            Fontsize of the title. The fontsizes of xlabel, ylabel,
+            tick_params, and legend are resized with 0.85, 0.85, 0.75,
+            and 0.75 fraction of title fontsize, respectively.
+
+        grid : bool
+            Whether to show (x,y) grid on the plot.
+
+        legend: bool
+            Whether to show legend on the plot.
+
+        legendloc: string
+            Legend location.
+
+        xlabel : string or None
+            Xlabel of the plot.
+
+        ylabel : string or None
+            Ylabel of the plot.
+
+        title : string or None
+            Title of the plot.
+
+        yscale: string or None
+            Scale for y-axis (coefficients). Valid options are
+            "linear", "log", "symlog", "logit". More on:
+            https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.yscale.html    
+
+        bbox_to_anchor: tuple, list or None
+            Relative coordinates for legend location outside of the plot.        
+
+        save_path: string or None
+            The full or relative path to save the plot including the image format.
+            For example "myplot.png" or "../../myplot.pdf"
+
+        Returns None
+        """
+        coeff_path_plot(self,
+                        feature_names=feature_names,
+                        figsize=figsize,
+                        linestyle=linestyle,
+                        fontsize=fontsize,
+                        grid=grid,
+                        legend=legend,
+                        legendloc=legendloc,
+                        xlabel=xlabel,
+                        ylabel=ylabel,
+                        title=title,
+                        yscale=yscale,
+                        bbox_to_anchor=bbox_to_anchor,
+                        save_path=save_path
+                       )
+        return None
+
+    def plot_cv_score(self, figsize=(10, 6), linestyle="--", linewidth=3, colors=None,
+                      marker="o", markersize=5, fontsize=18, grid=True, legend=True,
+                      legendloc="best", xlabel=None, ylabel=None, title=None, save_path=None
+                     ):
+        """Plot n-folds cross-validation scores vs -Log(lambda).
+
+        Parameters
+        ----------
+        figsize : tuple or list, as (width, height)
+            Figure size.
+
+        linestyle: string
+            Linestyle of the vertical lamda lines.
+
+        linewidth: integer or float
+            Linewidth of the vertical lamda lines.
+
+        colors: list or tuple
+            Colors of the marker, errorbar line, max_lambda line,
+            and best_lambda line, respectively. The default colors
+            are ("red", "black", "cyan", "blue"). The length of the
+            passed tuple/list should be always four.
+
+        marker: str, optional, (default="o")
+            Marker style. More details on:
+            (https://matplotlib.org/2.1.1/api/markers_api.html#module-matplotlib.markers)
+
+        markersize: intger or float
+            Markersize.
+
+        fontsize : int, float
+            Fontsize of the title. The fontsizes of xlabel, ylabel,
+            tick_params, and legend are resized with 0.85, 0.85, 0.75,
+            and 0.85 fraction of title fontsize, respectively.
+
+        grid : bool
+            Whether to show (x,y) grid on the plot.
+
+        legend: bool
+            Whether to show legend on the plot.
+
+        legendloc: string
+            Legend location.
+
+        xlabel : string or None
+            Xlabel of the plot.
+
+        ylabel : string or None
+            Ylabel of the plot.
+
+        title : string or None
+            Title of the plot.
+
+        save_path: string or None
+            The full or relative path to save the image including the image format.
+            For example "myplot.png" or "../../myplot.pdf"
+
+        Returns None
+        """
+        cv_score_plot(self,
+                      figsize=figsize,
+                      linestyle=linestyle,
+                      linewidth=linewidth,
+                      colors=colors,
+                      marker=marker,
+                      markersize=markersize,
+                      fontsize=fontsize,
+                      grid=grid,
+                      legend=legend,
+                      legendloc=legendloc,
+                      xlabel=xlabel,
+                      ylabel=ylabel,
+                      title=title,
+                      save_path=save_path)
+
+        return None

glmnet/logistic.py

@@ -16,6 +16,7 @@
                          _check_user_lambda,
                          _interpolate_model,
                          _score_lambda_path)
+from glmnet.plotting import coeff_path_plot, cv_score_plot
 
 
 class LogitNet(BaseEstimator):
@@ -203,6 +204,9 @@ def fit(self, X, y, sample_weight=None, relative_penalties=None, groups=None):
         self : object
             Returns self.
         """
+        # keep local copy of features
+        self.X_ = X
+
         X, y = check_X_y(X, y, accept_sparse='csr', ensure_min_samples=2)
         if sample_weight is None:
             sample_weight = np.ones(X.shape[0])
@@ -565,3 +569,158 @@ def score(self, X, y, lamb=None):
         """
         pred = self.predict(X, lamb=lamb)
         return np.apply_along_axis(accuracy_score, 0, pred, y)
+
+    def plot_coeff_path(self, feature_names=None, figsize=(10, 6), linestyle="-", fontsize=18,
+                        grid=True, legend=True, legendloc="center", xlabel=None, ylabel=None, 
+                        title=None, yscale=None, bbox_to_anchor=None, save_path=None):
+        """Plot coefficient's paths vs -Log(lambda).
+
+        Parameters
+        ----------
+        est : estimator
+            The previously fitted estimator.
+
+        feature_names : list, shape (n_features,)
+            Input features names neede for legend.
+
+        figsize : tuple or list, as (width, height)
+            Figure size.
+
+        linestyle: string
+            Linestyle of coefficients' paths.
+
+        fontsize : int, float
+            Fontsize of the title. The fontsizes of xlabel, ylabel,
+            tick_params, and legend are resized with 0.85, 0.85, 0.75,
+            and 0.75 fraction of title fontsize, respectively.
+
+        grid : bool
+            Whether to show (x,y) grid on the plot.
+
+        legend: bool
+            Whether to show legend on the plot.
+
+        legendloc: string
+            Legend location.
+
+        xlabel : string or None
+            Xlabel of the plot.
+
+        ylabel : string or None
+            Ylabel of the plot.
+
+        title : string or None
+            Title of the plot.
+
+        yscale: string or None
+            Scale for y-axis (coefficients). Valid options are
+            "linear", "log", "symlog", "logit". More on:
+            https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.yscale.html    
+
+        bbox_to_anchor: tuple, list or None
+            Relative coordinates for legend location outside of the plot.        
+
+        save_path: string or None
+            The full or relative path to save the plot including the image format.
+            For example "myplot.png" or "../../myplot.pdf"
+
+        Returns None
+        """
+        coeff_path_plot(self,
+                        feature_names=feature_names,
+                        figsize=figsize,
+                        linestyle=linestyle,
+                        fontsize=fontsize,
+                        grid=grid,
+                        legend=legend,
+                        legendloc=legendloc,
+                        xlabel=xlabel,
+                        ylabel=ylabel,
+                        title=title,
+                        yscale=yscale,
+                        bbox_to_anchor=bbox_to_anchor,
+                        save_path=save_path
+                       )
+        return None
+
+    def plot_cv_score(self, figsize=(10, 6), linestyle="--", linewidth=3, colors=None,
+                      marker="o", markersize=5, fontsize=18, grid=True, legend=True,
+                      legendloc="best", xlabel=None, ylabel=None, title=None, save_path=None
+                     ):
+        """Plot n-folds cross-validation scores vs -Log(lambda).
+
+        Parameters
+        ----------
+        figsize : tuple or list, as (width, height)
+            Figure size.
+
+        linestyle: string
+            Linestyle of the vertical lamda lines.
+
+        linewidth: integer or float
+            Linewidth of the vertical lamda lines.
+
+        colors: list or tuple
+            Colors of the marker, errorbar line, max_lambda line,
+            and best_lambda line, respectively. The default colors
+            are ("red", "black", "cyan", "blue"). The length of the
+            passed tuple/list should be always four.
+
+        marker: str, optional, (default="o")
+            Marker style. More details on:
+            (https://matplotlib.org/2.1.1/api/markers_api.html#module-matplotlib.markers)
+
+        markersize: intger or float
+            Markersize.
+
+        fontsize : int, float
+            Fontsize of the title. The fontsizes of xlabel, ylabel,
+            tick_params, and legend are resized with 0.85, 0.85, 0.75,
+            and 0.85 fraction of title fontsize, respectively.
+
+        grid : bool
+            Whether to show (x,y) grid on the plot.
+
+        legend: bool
+            Whether to show legend on the plot.
+
+        legendloc: string
+            Legend location.
+
+        xlabel : string or None
+            Xlabel of the plot.
+
+        ylabel : string or None
+            Ylabel of the plot.
+
+        title : string or None
+            Title of the plot.
+
+        save_path: string or None
+            The full or relative path to save the image including the image format.
+            For example "myplot.png" or "../../myplot.pdf"
+
+        Returns None
+        """
+        cv_score_plot(self,
+                      figsize=figsize,
+                      linestyle=linestyle,
+                      linewidth=linewidth,
+                      colors=colors,
+                      marker=marker,
+                      markersize=markersize,
+                      fontsize=fontsize,
+                      grid=grid,
+                      legend=legend,
+                      legendloc=legendloc,
+                      xlabel=xlabel,
+                      ylabel=ylabel,
+                      title=title,
+                      save_path=save_path)
+
+        return None
+
+
+
+
+