Added feature extraction class example

README.ipynb

@@ -106,7 +106,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Initialize image transformer."
+    "Initialize image transformer. There are three built-in feature extraction options, 'tsne', 'pca', and 'kpca' to align with the original MatLab implementation."
    ]
   },
   {
@@ -120,6 +120,27 @@
     "                      n_jobs=-1)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Alternatively, any class instance with method `.fit_transform()` that returns a 2-dimensional array of extracted features can also be provided to the ImageTransformer class. This allows for customization of the feature extraction procedure."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from sklearn.manifold import TSNE\n",
+    "\n",
+    "tsne = TSNE(n_components=2, perplexity=30, metric='cosine',\n",
+    "            random_state=1701, n_jobs=-1)\n",
+    "\n",
+    "it = ImageTransformer(feature_extractor=tsne, pixels=50)\n"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -131,7 +152,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 6,
    "metadata": {},
    "outputs": [
     {
@@ -145,21 +166,11 @@
       "needs_background": "light"
      },
      "output_type": "display_data"
-    },
-    {
-     "data": {
-      "text/plain": [
-       "<pyDeepInsight.image_transformer.ImageTransformer at 0x7f7549a01220>"
-      ]
-     },
-     "execution_count": 5,
-     "metadata": {},
-     "output_type": "execute_result"
     }
    ],
    "source": [
     "plt.figure(figsize=(5, 5))\n",
-    "it.fit(X_train_norm, plot=True)"
+    "_ = it.fit(X_train_norm, plot=True)"
    ]
   },
   {
@@ -171,7 +182,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 7,
    "metadata": {},
    "outputs": [
     {
@@ -211,7 +222,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": 8,
    "metadata": {},
    "outputs": [
     {
@@ -258,7 +269,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": 9,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -274,7 +285,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 9,
+   "execution_count": 10,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -291,7 +302,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": 11,
    "metadata": {},
    "outputs": [
     {
@@ -325,7 +336,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 11,
+   "execution_count": 12,
    "metadata": {},
    "outputs": [
     {

README.md

@@ -64,7 +64,7 @@ X_train_norm = ln.fit_transform(X_train)
 X_test_norm = ln.transform(X_test)
 ```
 
-Initialize image transformer.
+Initialize image transformer. There are three built-in feature extraction options, 'tsne', 'pca', and 'kpca' to align with the original MatLab implementation.
 
 
 ```python
@@ -73,29 +73,35 @@ it = ImageTransformer(feature_extractor='tsne',
                       n_jobs=-1)
 ```
 
+Alternatively, any class instance with method `.fit_transform()` that returns a 2-dimensional array of extracted features can also be provided to the ImageTransformer class. This allows for customization of the feature extraction procedure.
+
+
+```python
+from sklearn.manifold import TSNE
+
+tsne = TSNE(n_components=2, perplexity=30, metric='cosine',
+            random_state=1701, n_jobs=-1)
+
+it = ImageTransformer(feature_extractor=tsne, pixels=50)
+
+```
+
 Train image transformer on training data. Setting plot=True results in at 
 a plot showing the reduced features (blue points), convex full (red), and 
 minimum bounding rectagle (green) prior to rotation.
 
 
 ```python
 plt.figure(figsize=(5, 5))
-it.fit(X_train_norm, plot=True)
+_ = it.fit(X_train_norm, plot=True)
 ```
 
 
 
-![png](README_files/README_10_0.png)
+![png](README_files/README_12_0.png)
 
 
 
-
-
-
-    <pyDeepInsight.image_transformer.ImageTransformer at 0x7f7549a01220>
-
-
-
 The feature density matrix can be extracted from the trained transformer in order to view overall feature overlap.
 
 
@@ -116,7 +122,7 @@ _ = plt.title("Genes per pixel")
 
 
 
-![png](README_files/README_12_0.png)
+![png](README_files/README_14_0.png)
 
 
 
@@ -146,7 +152,7 @@ it.pixels = 50
 
 
 
-![png](README_files/README_14_0.png)
+![png](README_files/README_16_0.png)
 
 
 
@@ -179,7 +185,7 @@ plt.tight_layout()
 
 
 
-![png](README_files/README_20_0.png)
+![png](README_files/README_22_0.png)
 
 
 
@@ -199,7 +205,7 @@ plt.tight_layout()
 
 
 
-![png](README_files/README_22_0.png)
+![png](README_files/README_24_0.png)