updating figs

Author: carsen-stringer

paper/fig1.ipynb

@@ -16,7 +16,6 @@
     "from scipy.stats import zscore\n",
     "from rastermap.utils import bin1d\n",
     "\n",
-    "\n",
     "sys.path.insert(0, '/github/rastermap/paper/')\n",
     "import metrics, simulations, fig1\n",
     "\n",
@@ -37,16 +36,14 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "\n",
     "n_per_module = 1000\n",
     "for random_state in range(0, 10):\n",
     "    out = simulations.make_full_simulation(n_per_module=n_per_module, random_state=random_state)\n",
     "    spks, xi_all, stim_times_all, psth, psth_spont, iperm = out\n",
     "    np.savez(os.path.join(root, \"simulations/\", f\"sim_{random_state}.npz\"), \n",
     "                spks=spks, xi_all=xi_all, \n",
     "                stim_times_all=np.array(stim_times_all, dtype=object), \n",
-    "                psth=psth, psth_spont=psth_spont, iperm=iperm)\n",
-    "\n"
+    "                psth=psth, psth_spont=psth_spont, iperm=iperm)"
    ]
   },
   {
@@ -62,8 +59,6 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "import imp\n",
-    "imp.reload(simulations)\n",
     "simulations.embedding_performance(root, save=True)"
    ]
   },
@@ -82,7 +77,6 @@
    "source": [
     "# root path has folder \"simulations\" with saved results\n",
     "# will save figures to \"figures\" folder\n",
-    "imp.reload(fig1)\n",
     "os.makedirs(os.path.join(root, \"figures/\"), exist_ok=True)\n",
     "fig1.fig1(root, save_figure=True)   "
    ]
@@ -162,7 +156,6 @@
     "        M = metrics.run_UMAP(model.Usv, n_neighbors=nneigh)\n",
     "        embs_all[random_state, j] = M\n",
     "        j += 1\n",
-    "        print(j)\n",
     "    contamination_scores, triplet_scores = metrics.benchmarks(dat[\"xi_all\"], embs_all[random_state])\n",
     "    mnn, rho = metrics.embedding_quality_gt(dat[\"xi_all\"], embs_all[random_state], knn=knn.copy())\n",
     "    mnn_all[random_state], rho_all[random_state] = mnn, rho\n",
@@ -209,15 +202,14 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "imp.reload(fig1)\n",
     "fig1.suppfig_scores(root, save_figure=True)"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "### supp reproducible"
+    "### supp power-law only simulation"
    ]
   },
   {
@@ -226,15 +218,15 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "simulations.repro_algs(root)\n",
-    "fig1.suppfig_repro(root, save_fig=False)"
+    "# create sims and benchmark\n",
+    "simulations.spont_simulations(root)"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "### supp power-law only simulation"
+    "### supp reproducible"
    ]
   },
   {
@@ -243,10 +235,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# create sims and benchmark\n",
-    "simulations.spont_simulations(root)\n",
-    "# make figure\n",
-    "fig1.suppfig_spont(root)"
+    "simulations.repro_algs(root)\n",
+    "fig1.suppfig_repro(root, save_fig=True)"
    ]
   },
   {
@@ -267,6 +257,67 @@
     "# make figure\n",
     "fig1.suppfig_params(root)"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### supp no power-law noise added"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "\n",
+    "n_per_module = 1000\n",
+    "for random_state in range(1, 10):\n",
+    "    out = simulations.make_full_simulation(n_per_module=n_per_module, \n",
+    "                                            random_state=random_state, add_spont=False)\n",
+    "    spks, xi_all, stim_times_all, psth, psth_spont, iperm = out\n",
+    "    np.savez(os.path.join(root, \"simulations/\", f\"sim_no_add_spont_{random_state}.npz\"), \n",
+    "                spks=spks, xi_all=xi_all, \n",
+    "                stim_times_all=np.array(stim_times_all, dtype=object), \n",
+    "                psth=psth, psth_spont=psth_spont, iperm=iperm)\n",
+    "\n",
+    "# 6000 neurons in simulation with 5 modules\n",
+    "from tqdm import trange\n",
+    "embs_all = np.zeros((10, 7, 6000, 1))\n",
+    "scores_all = np.zeros((10, 2, 8, 5))\n",
+    "algos = [\"rastermap\", \"tSNE\", \"UMAP\", \"isomap\", \"laplacian\\neigenmaps\", \"hierarchical\\nclustering\", \"PCA\"]\n",
+    "\n",
+    "for random_state in trange(10):\n",
+    "    path = os.path.join(root, \"simulations\", f\"sim_no_add_spont_{random_state}.npz\")\n",
+    "    dat = np.load(path, allow_pickle=True)\n",
+    "    spks = dat[\"spks\"]\n",
+    "    embs, model = simulations.run_algos(spks, time_lag_window=10, locality=0.8)\n",
+    "\n",
+    "    # benchmarks\n",
+    "    contamination_scores, triplet_scores = metrics.benchmarks(dat[\"xi_all\"], \n",
+    "                                                embs.copy())\n",
+    "    embs_all[random_state] = embs\n",
+    "    scores_all[random_state] = np.stack((contamination_scores, triplet_scores), \n",
+    "                                        axis=0)\n",
+    "    if random_state==0:\n",
+    "        xi_all = dat[\"xi_all\"]\n",
+    "\n",
+    "np.savez(os.path.join(root, \"simulations\", \"sim_no_add_spont_performance.npz\"), \n",
+    "                scores_all=scores_all, \n",
+    "                embs_all=embs_all,\n",
+    "                xi_all=xi_all)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# make figure\n",
+    "fig1.suppfig_spont(root)"
+   ]
   }
  ],
  "metadata": {