transfer_learning/agglomerative_CNN_clustering.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# agglomerative_CNN_clustering\n",
    "- In [CNN_transfer_learning](./CNN_transfer_learning), we used a pre-trained convolutional neural network to generate a new feature set to better represent the galaxy images. These features were then compressed by a principal component analysis.\n",
    "- In [kmeans_CNN_clustering](./kmeans_CNN_clustering.ipynb), we did a basic grouping into 10 clusters. Since there is not really a _good_ answer for how many clusters of galaxies we should have, so now lets try an algorithm that dictates the number of clusters.\n",
    "- In [affinity_propagation_CNN_clustering](./affinity_propagation_CNN_clustering.ipynb), we used affinity propagation to automatically generate the clusters (and hence number and sizes). Using the higher dimensionality parameter space (i.e. output from CNN compress by the pca), we find that the process results in very precise clusters! The visual similarity is striking, however, is clustering based on postion angle (i.e. which direction the galaxy is pointing), colour, spiral (or other) structure and size/footprint of the galaxy in the image. Connecting these clusters is the difficult part, since many may simply be the same catagory/type of galaxy but viewed in a different way (position angle or footprint).\n",
    "- To help aid this, we now look to hierarchical clustering (i.e. agglomerative - bottom-up) clustering to help define several levels of similarity (i.e, using ~175 clusters as a starting point)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import harry_plotter\n",
    "import itertools\n",
    "from matplotlib import rcParams, gridspec\n",
    "harry_plotter.default(rcParams)\n",
    "\n",
    "from sklearn.cluster import AgglomerativeClustering\n",
    "from scipy.cluster import hierarchy\n",
    "from scipy.cluster.hierarchy import dendrogram, linkage\n",
    "import cluster_evaluate\n",
    "\n",
    "# default dendrogram colour params\n",
    "cmap = plt.cm.coolwarm(np.linspace(0.3, 2, 10))\n",
    "hierarchy.set_link_color_palette([mpl.colors.rgb2hex(rgb[:3]) for rgb in cmap])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Loading set of features \n",
    "- Also weights in case we want to manipulate feature space to rank pca components that best describe the variance in the data\n",
    "- Several feature representations were generated, so selecting combination of output features from CNN and the number of features these were compressed to by the pca."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'./cluster_plots/agglomerative/CNN-37632_PCA-100/'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pca_feat = 100\n",
    "cnn_feat = 37632\n",
    "\n",
    "features = np.load('./features/CNN-' + str(cnn_feat) + '_PCA-' + str(pca_feat) + '_features.npy')\n",
    "weights = np.load('./features/CNN-' + str(cnn_feat) + '_PCA-' + str(pca_feat) + '_weights.npy')\n",
    "\n",
    "plot_path = './cluster_plots/agglomerative/CNN-' + str(cnn_feat) + '_PCA-' + str(pca_feat) +'/'\n",
    "plot_path"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Defining agglomerative model for dendrogram\n",
    "- Defining one model to build hierarchical visualisation (i.e. dendrogram) : k=2.\n",
    "- Grabbing example images for each of these clusters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# defining k=2 tree\n",
    "agglo_2 = AgglomerativeClustering(n_clusters=2,\n",
    "                                  compute_distances=True,\n",
    "                                  linkage='ward')\n",
    "agglo_2.fit_predict(features) ;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# loading in pre-processed data\n",
    "data_dir = '../../sdss_CNN/data/cleaned_datasets/'\n",
    "x = np.load(data_dir + 'x.npy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "n_leaf_images = 16\n",
    "span = 10\n",
    "\n",
    "fig = plt.figure(figsize=(15, 5))\n",
    "gs = gridspec.GridSpec(span, n_leaf_images)\n",
    "\n",
    "ax_dendro = fig.add_subplot(gs[5:span, :])\n",
    "\n",
    "# plotting the final two images (i.e. representative for 2 clusters)\n",
    "ax_tt = []\n",
    "ax_tt.append(fig.add_subplot(gs[0:4, :int(n_leaf_images / 2)]) )\n",
    "ax_tt.append(fig.add_subplot(gs[0:4, int(n_leaf_images / 2):]) )\n",
    "ax_tt = np.array(ax_tt)\n",
    "\n",
    "for i, ax in enumerate(ax_tt):\n",
    "    ax.imshow(x[np.argmax(agglo_2.labels_ == i)])\n",
    "    ax.axis('off')\n",
    "\n",
    "# plotting structure of dendrogram\n",
    "param = cluster_evaluate.plot_dendrogram(agglo_2, \n",
    "                                        truncate_mode='level', p=3,\n",
    "                                        ax = ax_dendro,\n",
    "                                        color_threshold = 5, above_threshold_color = 'k',\n",
    "                                        leaf_font_size = 12, leaf_rotation=0,\n",
    "                                        orientation='top',\n",
    "                                        count_sort=True, \n",
    "                                        show_leaf_counts = True)\n",
    "#ax_dendro.axis('off')\n",
    "ax_dendro.axes.get_yaxis().set_visible(False)\n",
    "ax_dendro.spines['top'].set_visible(False)\n",
    "ax_dendro.spines['bottom'].set_visible(False)\n",
    "ax_dendro.spines['left'].set_visible(False)\n",
    "ax_dendro.spines['right'].set_visible(False)\n",
    "#ax_dendro.set_xlabel('Number of leaves (original samples) in branch', fontsize = 20)\n",
    "\n",
    "fig.subplots_adjust(wspace = 0.1, hspace= -0.5)\n",
    "plt.savefig(plot_path+'dendrogram-n16.png', format='png', dpi=200, bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Creating n=32 agglomerative clustering to save for output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# defining k=2 tree\n",
    "agglo_32 = AgglomerativeClustering(n_clusters=32,\n",
    "                                   linkage='ward')\n",
    "agglo_32.fit_predict(features) ;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure()\n",
    "ax = fig.add_subplot()\n",
    "\n",
    "sns.countplot(x=agglo_32.labels_, ax=ax)\n",
    "ax.set_xlabel('Cluster', fontsize=14)\n",
    "ax.set_ylabel('Galaxy count', fontsize=14)\n",
    "plt.savefig(plot_path + 'cluster_countplot.png', format='png', \n",
    "            bbox_inches='tight', dpi=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "for cl in np.unique(agglo_32.labels_):\n",
    "    cluster_evaluate.save_cluster_plot(images = x[agglo_32.labels_ == cl], \n",
    "                                       cluster_number = cl,\n",
    "                                       cluster_total= agglo_32.n_clusters - 1,\n",
    "                                       plot_path= plot_path) ;\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Saving cluster labels to file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "label_path = './cluster_tags/'\n",
    "np.save(label_path + 'agglomerative-' + str(agglo_32.n_clusters) + '_CNN-' + str(cnn_feat) + '_PCA-' + str(pca_feat) + '.npy', \n",
    "        agglo_32.labels_)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}