Merge pull request #2 from godatadriven/dev

Merge dev into master

README.md

@@ -6,6 +6,13 @@ This is an implementation of the model described in the following paper:
 
 I have copied some of the code from the [paper's code base](https://github.com/elisim/piven), and cite the author's paper where this is the case.
 
+<figure>
+    <img src="https://github.com/elisim/piven/blob/master/piven_architecture.jpg" height=250 width=800>
+    <figcaption>
+        <i>NN with piven layer, from from Simhayev, Gilad and Rokach (2020).</i>
+    </figcaption>
+</figure>
+
 ## In short
 
 A neural network with a Piven (Prediction Intervals with specific value prediction) output layer returns a point
@@ -69,14 +76,14 @@ This will save the metrics, keras model, and model predictions to the folder.
 
 For additional examples, see the 'tests' and 'notebooks' folders.
 
-## Creating your own Piven model
-
-You can use a Piven layer on any neural network architecture. REF REPO AUTHORS example convo network.
+## Creating your own model with Piven layer
 
-This library uses the KerasRegressor wrapper from the tensorflow library to make scikit-compatible
-keras models.  
+You can use a Piven layer on any neural network architecture. The authors of the Piven paper use it on top of
+a bunch of [CNN layers](https://github.com/elisim/piven/blob/master/imdb/main.py) to predict people's age.
 
-Suppose that you want to create an Model with a Piven output layer. You would first specify a build
+Suppose that you want to create an Model with a Piven output layer. Because this module uses the 
+[KerasRegressor](https://www.tensorflow.org/api_docs/python/tf/keras/wrappers/scikit_learn/KerasRegressor)  wrapper 
+from the tensorflow library to make scikit-compatible keras models, you would first specify a build
 function like so:
 
 ```python
@@ -120,7 +127,7 @@ class MyPivenModel(PivenBaseModel):
 
     @classmethod
     def load(cls, path: str):
-        Model_config = MyPivenModel.load_Model_config(path)
+        model_config = MyPivenModel.load_model_config(path)
         model = MyPivenModel.load_model_from_disk(build_fn=piven_model, path)
         run = cls(**Model_config)
         run.model = model
@@ -141,5 +148,3 @@ Note that the inputs to `MyPivenModel` must match the inputs to the `piven_model
 You can now call all methods defined as in the PivenBaseModel class. Check the 
 [PivenMlpModel class](https://gitlab.com/jasperginn/piven.py/-/blob/dev/src/piven/Models/mlp_regressor.py)
 for a more detailed example.
-
-

notebooks/year_prediction_msd.ipynb

@@ -9,7 +9,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -18,7 +18,7 @@
     "import numpy as np\n",
     "import matplotlib.pyplot as plt\n",
     "import seaborn as sns\n",
-    "from piven.experiments import PivenMlpExperiment\n",
+    "from piven.models import PivenMlpModel\n",
     "import tensorflow as tf\n",
     "from sklearn.compose import TransformedTargetRegressor\n",
     "\n",
@@ -27,7 +27,41 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": null,
+   "metadata": {
+    "pycharm": {
+     "name": "#%%\n"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "records = []\n",
+    "with dpath.open(\"r\") as infile:\n",
+    "    for line in infile.readlines():\n",
+    "        line_split = line.replace(\"\\n\", \"\").split(\",\")\n",
+    "        year = int(line_split[0])\n",
+    "        inputs = [*map(lambda x: float(x), line_split[1:])]\n",
+    "        records.append([year] + inputs)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "records = []\n",
+    "with dpath.open(\"r\") as infile:\n",
+    "    for line in infile.readlines():\n",
+    "        line_split = line.replace(\"\\n\", \"\").split(\",\")\n",
+    "        year = int(line_split[0])\n",
+    "        inputs = [*map(lambda x: float(x), line_split[1:])]\n",
+    "        records.append([year] + inputs)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -1620,9 +1654,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "PyCharm (pivenregressor)",
    "language": "python",
-   "name": "python3"
+   "name": "pycharm-b9238b0d"
   },
   "language_info": {
    "codemirror_mode": {

src/piven/models/base.py

@@ -92,7 +92,7 @@ def score(
         }
 
     @staticmethod
-    def load_experiment_config(path: str):
+    def load_model_config(path: str):
         if not (Path(path) / "experiment_params.json").is_file():
             raise FileNotFoundError(f"No experiment file found in {path}.")
         with (Path(path) / "experiment_params.json").open("r") as infile:

src/piven/models/mlp_regressor.py

@@ -87,8 +87,8 @@ def build_model(self, preprocess: Union[None, Pipeline, TransformerMixin] = None
 
     @classmethod
     def load(cls, path: str):
-        experiment_config = PivenMlpModel.load_experiment_config(path)
+        model_config = PivenMlpModel.load_model_config(path)
         model = PivenMlpModel.load_model_from_disk(piven_model, path)
-        run = cls(**experiment_config)
+        run = cls(**model_config)
         run.model = model
         return run