notebooks/tutorials/tutorial-evaluate-run-time-gp.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "practical-attitude",
   "metadata": {},
   "source": [
    "# Evaluate running time by number of samples"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "68015dab-05de-4856-9548-c450cd94685d",
   "metadata": {},
   "source": [
    "In this tutorial we evaluate the running time of ternadecov for a varying number of samples using the GP trajectory model. The configuration of this tutorial has been set so that it completes in under 5 minutes. In order to generate more interesting plots increase the `n_iters` to 20k and set `n_sample_range` to span the range 500-1000."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "educational-rochester",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load libraries\n",
    "import anndata\n",
    "import matplotlib.pylab as plt\n",
    "import torch\n",
    "from ternadecov.time_deconv import *\n",
    "from ternadecov.simulator import *\n",
    "from ternadecov.stats_helpers import *\n",
    "from ternadecov.dataset import *\n",
    "from ternadecov.deconvolution_plotter import *\n",
    "from ternadecov.parametrization import *\n",
    "from ternadecov.evaluation import evaluate_with_trajectory\n",
    "from ternadecov.deconvolution_writer import DeconvolutionWriter\n",
    "from ternadecov.deconvolution_exporter import DeconvolutionExporter"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "blond-cycle",
   "metadata": {},
   "source": [
    "## Configure"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "a3197af2-17d1-4bc0-80b4-20ede3fdeffc",
   "metadata": {},
   "outputs": [],
   "source": [
    "# number of iterations to run\n",
    "n_iters = 5_000\n",
    "\n",
    "# path to single-cell anndata dataset\n",
    "sc_anndata_path = \"/home/nbarkas/disk2/deconvolution_method/datasets/ebov/load_data_python/ebov_sc.h5ad\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "south-speech",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(sc_anndata_path, 'rb') as fh:\n",
    "    sc_anndata = anndata.read_h5ad(fh)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "caroline-while",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Simulate data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "3a235622-8acd-4ff7-aba2-e3af8d985726",
   "metadata": {},
   "outputs": [],
   "source": [
    "types = DeconvolutionDatatypeParametrization()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4cf9734-729a-4344-8c99-3bfd08f975de",
   "metadata": {},
   "source": [
    "Next we generate a 'SingleCellDataset' object. This is a special type of deconvolution dataset object that only contains single-cell data and is used only for simulations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "iraqi-difference",
   "metadata": {},
   "outputs": [],
   "source": [
    "ebov_sc_dataset = SingleCellDataset(\n",
    "    sc_anndata, \n",
    "    sc_celltype_col=\"Subclustering_reduced\", \n",
    "    dtype_np=types.dtype_np, dtype = types.dtype, device=types.device)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d3eaffb7-0b6b-43ca-89ac-8091aef721b7",
   "metadata": {},
   "source": [
    "Sample a random trajectory of celltypes that will be used for all iterations by only varying the number of samples draw from it\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "first-october",
   "metadata": {},
   "outputs": [],
   "source": [
    "trajectory_coef = sample_trajectories(\n",
    "    type = 'periodic',\n",
    "    num_cell_types = ebov_sc_dataset.w_hat_gc.shape[1]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "c82765a8-756b-4724-b2ea-084ce3cbc598",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[10, 30, 50, 70, 100, 300, 500, 700, 900]\n"
     ]
    }
   ],
   "source": [
    "n_sample_range = list(range(10,90,20))+list(range(100,1000,200)) \n",
    "print(n_sample_range)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "respected-spirituality",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.44106e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2983 genes selected\n",
      "[step: 0, time: 1 s ] loss: 318857.12\n",
      "[step: 1000, time: 14 s ] loss: 232587.94\n",
      "[step: 2000, time: 27 s ] loss: 189270.65\n",
      "[step: 3000, time: 41 s ] loss: 172177.17\n",
      "[step: 4000, time: 54 s ] loss: 166771.14\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.2013e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3056 genes selected\n",
      "[step: 0, time: 1 s ] loss: 987292.43\n",
      "[step: 1000, time: 16 s ] loss: 689260.68\n",
      "[step: 2000, time: 32 s ] loss: 560391.30\n",
      "[step: 3000, time: 47 s ] loss: 507137.19\n",
      "[step: 4000, time: 63 s ] loss: 488356.32\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.33856e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3105 genes selected\n",
      "[step: 0, time: 1 s ] loss: 1607024.61\n",
      "[step: 1000, time: 19 s ] loss: 1143708.06\n",
      "[step: 2000, time: 37 s ] loss: 930685.83\n",
      "[step: 3000, time: 55 s ] loss: 839948.73\n",
      "[step: 4000, time: 74 s ] loss: 806818.78\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.31587e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3143 genes selected\n",
      "[step: 0, time: 1 s ] loss: 2336701.00\n",
      "[step: 1000, time: 21 s ] loss: 1637840.84\n",
      "[step: 2000, time: 42 s ] loss: 1313826.42\n",
      "[step: 3000, time: 62 s ] loss: 1186340.73\n",
      "[step: 4000, time: 83 s ] loss: 1140662.78\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.27915e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3176 genes selected\n",
      "[step: 0, time: 1 s ] loss: 3449919.09\n",
      "[step: 1000, time: 26 s ] loss: 2377165.59\n",
      "[step: 2000, time: 51 s ] loss: 1900222.40\n",
      "[step: 3000, time: 76 s ] loss: 1712169.05\n",
      "[step: 4000, time: 101 s ] loss: 1645778.51\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.29993e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3251 genes selected\n",
      "[step: 0, time: 1 s ] loss: 10407189.38\n",
      "[step: 1000, time: 78 s ] loss: 7202534.87\n",
      "[step: 2000, time: 156 s ] loss: 5788878.75\n",
      "[step: 3000, time: 233 s ] loss: 5211359.14\n",
      "[step: 4000, time: 312 s ] loss: 4998428.71\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.32307e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3260 genes selected\n",
      "[step: 0, time: 1 s ] loss: 17112167.40\n",
      "[step: 1000, time: 116 s ] loss: 11974210.10\n",
      "[step: 2000, time: 232 s ] loss: 9664693.06\n",
      "[step: 3000, time: 347 s ] loss: 8702128.97\n",
      "[step: 4000, time: 464 s ] loss: 8351270.65\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.30408e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3249 genes selected\n",
      "[step: 0, time: 1 s ] loss: 23947811.63\n",
      "[step: 1000, time: 207 s ] loss: 16772669.62\n",
      "[step: 2000, time: 413 s ] loss: 13486545.63\n",
      "[step: 3000, time: 620 s ] loss: 12153806.41\n",
      "[step: 4000, time: 829 s ] loss: 11650319.40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/anndata/_core/anndata.py:120: ImplicitModificationWarning: Transforming to str index.\n",
      "  warnings.warn(\"Transforming to str index.\", ImplicitModificationWarning)\n",
      "/home/nbarkas/disk1/anaconda3/lib/python3.9/site-packages/sklearn/linear_model/_ridge.py:147: LinAlgWarning: Ill-conditioned matrix (rcond=5.30656e-10): result may not be accurate.\n",
      "  return linalg.solve(A, Xy, sym_pos=True,\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3298 genes selected\n",
      "[step: 0, time: 1 s ] loss: 30952238.69\n",
      "[step: 1000, time: 300 s ] loss: 21739736.82\n",
      "[step: 2000, time: 600 s ] loss: 17492538.29\n",
      "[step: 3000, time: 900 s ] loss: 15769364.36\n",
      "[step: 4000, time: 1203 s ] loss: 15126894.97\n"
     ]
    }
   ],
   "source": [
    "n_samples = []\n",
    "l1_error_norm = []\n",
    "fit_time = []\n",
    "\n",
    "\n",
    "for n in n_sample_range:\n",
    "    \n",
    "    eval_result = evaluate_with_trajectory(\n",
    "        sc_dataset = ebov_sc_dataset, \n",
    "        n_samples = n, \n",
    "        trajectory_type = 'periodic', \n",
    "        trajectory_coef = trajectory_coef,\n",
    "        n_iters = n_iters,\n",
    "        types = DeconvolutionDatatypeParametrization(),\n",
    "        deconvolution_params = {\n",
    "            'trajectory_model_type': 'gp',\n",
    "            'hyperparameters': TimeRegularizedDeconvolutionModelParametrization(),\n",
    "            'trajectory_hyperparameters': TimeRegularizedDeconvolutionGPParametrization(),\n",
    "        }\n",
    "    )\n",
    "    \n",
    "    n_samples.append(eval_result['n_samples'])\n",
    "    l1_error_norm.append(eval_result['l1_error_norm'])\n",
    "    fit_time.append(eval_result['fit_time'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "7d5cbd22-e19a-4c30-bfa9-a1e8c0bc7045",
   "metadata": {},
   "outputs": [],
   "source": [
    "evaluation_resuts_df = pd.DataFrame({'n_samples':n_samples, 'l1_error_norm': list(x.item() for x in l1_error_norm), 'fit_time': fit_time})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "2ddaa13d-7986-43b4-b6c7-bffce08c7429",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>n_samples</th>\n",
       "      <th>l1_error_norm</th>\n",
       "      <th>fit_time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>10</td>\n",
       "      <td>0.871758</td>\n",
       "      <td>66.451277</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>30</td>\n",
       "      <td>0.869518</td>\n",
       "      <td>78.291943</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>50</td>\n",
       "      <td>0.864875</td>\n",
       "      <td>91.768553</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>70</td>\n",
       "      <td>0.850476</td>\n",
       "      <td>104.141044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>100</td>\n",
       "      <td>0.859508</td>\n",
       "      <td>126.601906</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>300</td>\n",
       "      <td>0.852567</td>\n",
       "      <td>390.698328</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>500</td>\n",
       "      <td>0.854624</td>\n",
       "      <td>582.720403</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>700</td>\n",
       "      <td>0.848973</td>\n",
       "      <td>1039.847996</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>900</td>\n",
       "      <td>0.853655</td>\n",
       "      <td>1507.635575</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   n_samples  l1_error_norm     fit_time\n",
       "0         10       0.871758    66.451277\n",
       "1         30       0.869518    78.291943\n",
       "2         50       0.864875    91.768553\n",
       "3         70       0.850476   104.141044\n",
       "4        100       0.859508   126.601906\n",
       "5        300       0.852567   390.698328\n",
       "6        500       0.854624   582.720403\n",
       "7        700       0.848973  1039.847996\n",
       "8        900       0.853655  1507.635575"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "evaluation_resuts_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "2910865d-0ace-4bd5-a0e6-edba8a8fc8ab",
   "metadata": {},
   "outputs": [],
   "source": [
    "evaluation_resuts_df.to_csv('runtime_evaluation_results.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9a70a509-00ef-4eb5-85f7-ffd1508c4752",
   "metadata": {},
   "source": [
    "## Plot results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "52d12135-44a4-45b6-844b-629e83b8e22e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot run-time\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.rcParams.update({'font.size': 22})\n",
    "plt.plot(evaluation_resuts_df.n_samples.to_numpy(), evaluation_resuts_df.fit_time.to_numpy(), color='black')\n",
    "plt.plot([10],[3474], color='red', markersize=10, marker='o')\n",
    "plt.xlabel(\"Number of bulk samples\")\n",
    "plt.ylabel(\"Time (s)\")\n",
    "plt.title(\"ternadecov run time\")\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "environment": {
   "name": "pytorch-gpu.1-7.m65",
   "type": "gcloud",
   "uri": "gcr.io/deeplearning-platform-release/pytorch-gpu.1-7:m65"
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}