Merge branch 'development' of github.com:gjbex/Python-for-machine-learning into development

Author: gjbex

environment.yml

@@ -1,6 +1,6 @@
 name: python_for_machine_learning
 channels:
-  - defaults
+  - conda-forge
 dependencies:
   - keras
   - scikit-learn
@@ -11,6 +11,8 @@ dependencies:
   - seaborn
   - holoviews
   - ipywidgets
-    #  - hyperopt
-prefix: /home/gjb/miniconda3/envs/python_for_machine_learning
-
+  - lime
+  - ca-certificates
+  - certifi
+  - openssl
+prefix: /home/gjb/mambaforge/envs/python_for_machine_learning

python_for_machine_learning_linux64_conda_specs.txt

@@ -30,3 +30,5 @@ to create it.  There is some material not covered in the presentation as well.
   the Kullback-Leibler divergence for probability distributions.
 * [`curse_of_dimensionality.ipynb`](curse_of_dimensionality.ipynb): illustration of
   the curse of dimensionality.
+* [`lime`](lime/): illustration of using LIME to explain machine learning
+  models.

source-code/README.md

@@ -0,0 +1,12 @@
+# LIME
+
+LIME is a tool to make machine learning models explainable.  It is a blackbox
+algorithm that simply requires a function that will generate the output of the
+model for given input.
+
+
+## What is it?
+
+1. `wine_and_lime.ipynb`: application of LIME to a random forest classification
+   model for a Kaggle dataset on the quality of white wine.
+1. `data`: directory containing the data.