visualize_topic_decompostion

dev/Grad_norm.py → dev/Grad_norm_dev.py

@@ -8,7 +8,7 @@
 
 
 class Grad_Norm:
-    def __init__(self, X=None, Z=None, sigma=None, resolution=None,
+    def __init__(self, X=None, Z=None, Y=None, sigma=None, resolution=None,
                  labels=None, fig_size=[6, 6], title_text='Grad_norm', cmap_type='jet',
                  interpolation_method='spline36', repeat=False, interval=5):
         # インプットが無効だった時のエラー処理

dev/fetch_make_csv.ipynb

@@ -1346,10 +1346,134 @@
     "print(\"count:\", len(divs))"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "6b3ed84f-142e-42af-a8be-05b30e24c897",
+   "metadata": {},
+   "source": [
+    "こっから違う話"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 33,
+   "id": "b2424a1a-45f3-41f3-b7c0-f729425b2e03",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[ 0,  1,  2,  3],\n",
+       "       [ 4,  5,  6,  7],\n",
+       "       [ 8,  9, 10, 11]])"
+      ]
+     },
+     "execution_count": 33,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import numpy as np\n",
+    "\n",
+    "X = np.arange(12).reshape(3,4)\n",
+    "mask_std = np.zeros(X.shape)\n",
+    "X"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 34,
+   "id": "57eaab59-0d09-4197-be0a-ff7af634d692",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "[3 3 3]\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(np.argmax(X, axis=1))\n",
+    "mask = np.argmax(X, axis=1)\n",
+    "k_max = np.max(X, axis=0)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 35,
+   "id": "411d5d54-d642-41fe-8f5d-a7991a68db67",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[0.        , 0.        , 0.        , 0.27272727],\n",
+       "       [0.        , 0.        , 0.        , 0.63636364],\n",
+       "       [0.        , 0.        , 0.        , 1.        ]])"
+      ]
+     },
+     "execution_count": 35,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# For mask and normalization(min:0, max->1)\n",
+    "for i, max_k in enumerate(mask):\n",
+    "    mask_std[i, max_k] = 1 / k_max[max_k]\n",
+    "X_mask_std = X * mask_std\n",
+    "\n",
+    "\n",
+    "\n",
+    "    "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 45,
+   "id": "c194345f-857c-42de-b648-0ccd4fa5b1c5",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "['rgb(31, 119, 180, 0.1)',\n",
+       " 'rgb(255, 127, 14, 0.1)',\n",
+       " 'rgb(44, 160, 44, 0.1)',\n",
+       " 'rgb(214, 39, 40, 0.1)',\n",
+       " 'rgb(148, 103, 189, 0.1)',\n",
+       " 'rgb(140, 86, 75, 0.1)',\n",
+       " 'rgb(227, 119, 194, 0.1)',\n",
+       " 'rgb(127, 127, 127, 0.1)',\n",
+       " 'rgb(188, 189, 34, 0.1)',\n",
+       " 'rgb(23, 190, 207, 0.1)']"
+      ]
+     },
+     "execution_count": 45,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "DEFAULT_PLOTLY_COLORS=['rgb(31, 119, 180)', 'rgb(255, 127, 14)',\n",
+    "                       'rgb(44, 160, 44)', 'rgb(214, 39, 40)',\n",
+    "                       'rgb(148, 103, 189)', 'rgb(140, 86, 75)',\n",
+    "                       'rgb(227, 119, 194)', 'rgb(127, 127, 127)',\n",
+    "                       'rgb(188, 189, 34)', 'rgb(23, 190, 207)']\n",
+    "alpha = 0.1\n",
+    "# for k in DEFAULT_PLOTLY_COLORS:\n",
+    "#     print(\")\n",
+    "DPC_with_Alpha = [k[:-1]+', '+str(alpha)+k[-1:] for k in DEFAULT_PLOTLY_COLORS]\n",
+    "DPC_with_Alpha"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "cea8906c-059b-4514-a61a-878f5796fa11",
+   "id": "0a096ee8-2033-4b65-97f5-c777067c3a76",
    "metadata": {},
    "outputs": [],
    "source": []

fit.py

@@ -1,8 +1,11 @@
+# import sys
+# sys.path.append('../')
 from som import ManifoldModeling as MM
 import matplotlib.pyplot as plt
 import matplotlib.animation as ani
 import numpy as np
 import pickle
+from sklearn.decomposition import NMF
 
 def update(i, fig_title, Z):
     if i != 0:
@@ -12,7 +15,7 @@ def update(i, fig_title, Z):
 
 
 if __name__ == '__main__':
-    from dev.Grad_norm import Grad_Norm
+    from dev.Grad_norm_dev import Grad_Norm
 
     keyword = "ファッション"
     model = "SOM"
@@ -36,7 +39,7 @@ def update(i, fig_title, Z):
 
     np.random.seed(seed)
 
-    mm = MM(X, latent_dim=latent_dim, resolution=resolution, sigma_max=sigma_max, sigma_min=sigma_min, tau=tau, init='PCA')
+    mm = MM(X, model_name='SOM', latent_dim=latent_dim, resolution=resolution, sigma_max=sigma_max, sigma_min=sigma_min, tau=tau, init='PCA')
     mm.fit(nb_epoch=nb_epoch)
 
     mm_umatrix = Grad_Norm(X=X,
@@ -45,11 +48,23 @@ def update(i, fig_title, Z):
                             labels=labels,
                             resolution=umat_resolution,
                             title_text=title_text)
-    mm_umatrix.draw_umatrix()
+    # mm_umatrix.draw_umatrix()
+    Z = mm.history['z'][-1]
+    Y = mm.history['y'][-1]
+    model_t3 = NMF(n_components=5, init='random', random_state=2, max_iter=300,
+                           solver='cd')
+    Wt3 = model_t3.fit_transform(Y)
+    Ht3 = model_t3.components_
+
+    fig = plt.figure(figsize=(16, 9))
+    two = plt.imshow(Wt3[:, 1].reshape(resolution, resolution),
+                                   extent=[Z[:, 0].min(), Z[:, 0].max(), Z[:, 1].min(),
+                                           Z[:, 1].max()],
+                                   interpolation=None, alpha=0.8)
+    plt.show()
+
 
-    Fig = plt.figure(figsize=(16, 9))
     # animation = ani.FuncAnimation(Fig, update, fargs=(np.zeros(10, 10, 2)))
-    fig = plt.figure(figsize = (10, 6))
     # ani = ani.FuncAnimation(fig, update, fargs = ('Initial Animation! ', mm.history['z']), \
     # interval = 1, frames = 500)
     # plt.show()

som.py

@@ -131,5 +131,3 @@ def fit(self, nb_epoch=100, verbose=True):
             self.history['z'][epoch] = self.Z
             self.history['y'][epoch] = self.Y
             self.history['sigma'][epoch] = self.sigma
-
-