Upload main.ipynb with huggingface_hub

main.ipynb

@@ -911,99 +911,6 @@
     "plt.show()"
    ]
   },
-  {
-   "cell_type": "code",
-   "execution_count": 551,
-   "id": "39db4fa4-b188-4ee0-b626-fc08c69ab1e7",
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "100%|███████████████████████████████████████| 181/181 [00:00<00:00, 3041.59it/s]"
-     ]
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "tensor([ 0.0720, -0.1333,  0.7357,  ..., -0.4051, -0.2543, -0.2446])\n"
-     ]
-    },
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "\n"
-     ]
-    }
-   ],
-   "source": [
-    "# Calculate pairwise distances for paired indices\n",
-    "paired_distances = None\n",
-    "for pair in tqdm(pairs):\n",
-    "    idx1, idx2 = pair\n",
-    "    if paired_distances is None:\n",
-    "        paired_distances = torch.abs(vox[idx1] - vox[idx2])[None]\n",
-    "    else:\n",
-    "        paired_distances = torch.vstack((paired_distances, torch.abs(vox[idx1] - vox[idx2])[None]))\n",
-    "\n",
-    "# Calculate pairwise distances for unpaired indices\n",
-    "unpaired_distances = None\n",
-    "for i in range(len(vox)):\n",
-    "    for j in range(i + 1, len(vox)):\n",
-    "        if unpaired_distances is not None:\n",
-    "            if len(unpaired_distances) == len(paired_distances):\n",
-    "                break\n",
-    "        if [i, j] not in pairs and [j, i] not in pairs:\n",
-    "            if unpaired_distances is None:\n",
-    "                unpaired_distances = torch.abs(vox[i] - vox[j])[None]\n",
-    "            else:\n",
-    "                unpaired_distances = torch.vstack((unpaired_distances, torch.abs(vox[i] - vox[j])[None]))\n",
-    "\n",
-    "# Calculate mean distances\n",
-    "mean_paired_distance = torch.median(paired_distances, dim=0).values\n",
-    "mean_unpaired_distance = torch.median(unpaired_distances, dim=0).values\n",
-    "\n",
-    "# Calculate reliability metric for each voxel\n",
-    "reliability_metric = mean_unpaired_distance - mean_paired_distance\n",
-    "\n",
-    "print(reliability_metric)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 426,
-   "id": "b691ed58-0d87-42c8-94a0-f2bbe5b52620",
-   "metadata": {},
-   "outputs": [
-    {
-     "data": {
-      "image/png": "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",
-      "text/plain": [
-       "<Figure size 640x480 with 1 Axes>"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "tensor(68)\n"
-     ]
-    }
-   ],
-   "source": [
-    "plt.plot(np.sort(reliability_metric.cpu().numpy()))\n",
-    "plt.show()\n",
-    "\n",
-    "subset = reliability_metric>3\n",
-    "print(subset.sum())"
-   ]
-  },
   {
    "cell_type": "markdown",
    "id": "a8866ce2-1cf2-459e-aa81-dec26a3dcd33",