Merge pull request #2 from EpistasisLab/master

merge

.github/workflows/build-docs.yml

@@ -0,0 +1,18 @@
+name: build-docs
+on:
+  push:
+    branches:
+      - master
+    paths:
+      - docs_sources/**
+      - mkdocs.yml
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - uses: actions/setup-python@v2
+        with:
+          python-version: 3.x
+      - run: pip install mkdocs-material 
+      - run: mkdocs gh-deploy --force

docs_sources/api.md

@@ -93,7 +93,7 @@ Cross-validation strategy used when evaluating pipelines.
 <br /><br />
 Possible inputs:
 <ul>
-<li>integer, to specify the number of folds in a StratifiedKFold,</li>
+<li>integer, to specify the number of folds in an unshuffled StratifiedKFold,</li>
 <li>An object to be used as a cross-validation generator, or</li>
 <li>An iterable yielding train/test splits.</li>
 </blockquote>
@@ -601,7 +601,7 @@ Cross-validation strategy used when evaluating pipelines.
 <br /><br />
 Possible inputs:
 <ul>
-<li>integer, to specify the number of folds in a KFold,</li>
+<li>integer, to specify the number of folds in an unshuffled KFold,</li>
 <li>An object to be used as a cross-validation generator, or</li>
 <li>An iterable yielding train/test splits.</li>
 </ul>

docs_sources/using.md

@@ -722,3 +722,4 @@ A simple example of using TPOT-NN is shown in [examples](/tpot/examples/).
 - TPOT will occasionally learn pipelines that stack several `sklearn` estimators. Mathematically, these can be nearly identical to some deep learning models. For example, by stacking several `sklearn.linear_model.LogisticRegression`s, you end up with a very close approximation of a Multilayer Perceptron; one of the simplest and most well known deep learning architectures. TPOT's genetic programming algorithms generally optimize these 'networks' much faster than PyTorch, which typically uses a more brute-force convex optimization approach.
 
 - The problem of 'black box' model introspection is one of the most substantial criticisms and challenges of deep learning. This problem persists in `tpot.nn`, whereas TPOT's default estimators often are far easier to introspect.
+

mkdocs.yml

@@ -7,11 +7,32 @@ repo_url: https://github.com/epistasislab/tpot
 edit_uri: edit/master/docs_sources/
 docs_dir: docs_sources/
 site_dir: docs/
-theme: readthedocs
+  #theme: readthedocs
+theme:
+  name: material
+  features:
+    - toc.integrate
+  palette:
+    # light mode
+    - scheme: default
+      toggle:
+        icon: material/brightness-7
+        name: Switch to dark mode
+
+    # dark mode
+    - scheme: slate
+      toggle:
+        icon: material/brightness-4
+        name: Switch to light mode
 
 markdown_extensions:
   - tables
   - fenced_code
+  - pymdownx.highlight:
+      anchor_linenums: true
+  - pymdownx.inlinehilite
+  - pymdownx.snippets
+  - pymdownx.superfences
 
 copyright: Developed by <a href="http://www.randalolson.com">Randal S. Olson</a> and others at the University of Pennsylvania
 

tpot/base.py

@@ -1561,10 +1561,10 @@ def _evaluate_individuals(
                             ]
                         ]
 
-                        self.dask_graphs_ = tmp_result_scores
+
                         with warnings.catch_warnings():
                             warnings.simplefilter("ignore")
-                            tmp_result_scores = list(dask.compute(*tmp_result_scores))
+                            tmp_result_scores = list(dask.compute(*tmp_result_scores, num_workers=self.n_jobs))
 
                     else:
 
@@ -1812,7 +1812,7 @@ def _mate_operator(self, ind1, ind2):
                 offspring.statistics["generation"] = "INVALID"
                 break
 
-        return offspring, offspring2
+        return offspring, offspring2, self.evaluated_individuals_
 
     @_pre_test
     def _random_mutation_operator(self, individual, allow_shrink=True):

tpot/gp_deap.py

@@ -131,13 +131,17 @@ def varOr(population, toolbox, lambda_, cxpb, mutpb):
         if op_choice < cxpb:  # Apply crossover
             ind1, ind2 = pick_two_individuals_eligible_for_crossover(population)
             if ind1 is not None:
-                ind1, _ = toolbox.mate(ind1, ind2)
+                ind1_cx, _, evaluated_individuals_= toolbox.mate(ind1, ind2)
                 del ind1.fitness.values
+
+                if str(ind1_cx) in evaluated_individuals_:
+                    ind1_cx = mutate_random_individual(population, toolbox)
+                offspring.append(ind1_cx)
             else:
                 # If there is no pair eligible for crossover, we still want to
                 # create diversity in the population, and do so by mutation instead.
-                ind1 = mutate_random_individual(population, toolbox)
-            offspring.append(ind1)
+                ind_mu = mutate_random_individual(population, toolbox)
+                offspring.append(ind_mu)
         elif op_choice < cxpb + mutpb:  # Apply mutation
             ind = mutate_random_individual(population, toolbox)
             offspring.append(ind)