change factors for data pyramid

holoviz-topics · Sep 28, 2024 · 0a2b9c2 · 0a2b9c2
1 parent b332094
commit 0a2b9c2
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 28 deletions.
diff --git a/multichannel_timeseries/assets/.ipynb_checkpoints/240927_multichan-checkpoint.png b/multichannel_timeseries/assets/.ipynb_checkpoints/240927_multichan-checkpoint.png
diff --git a/multichannel_timeseries/index.ipynb b/multichannel_timeseries/index.ipynb
@@ -20,21 +20,12 @@
     "___"
    ]
   },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "#TODO: Make Key Features list into a diagram or gif showing the feature-components of the viewer"
-   ]
-  },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
     "### Key Features\n",
-    "Analyzing electrophysiological data often involves searching for patterns across time, channels, and covariates. Features that support this type of investigation for time-aligned, amplitude-diverse data include:\n",
+    "Analyzing electrophysiological data often involves searching for patterns across time, channels, and covariates. Target features that support this type of investigation for time-aligned, amplitude-diverse data include:\n",
     "\n",
     "- **Smooth Interactions at Scale:** Smooth zooming and panning across time and channels.\n",
     "- **Subcoordinate Axes:** Independent amplitude dimension (y-axis) per channel.\n",
@@ -63,7 +54,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "<img src='./assets/240716b_multi-chan-ts.png' alt=\"Multichannel timeseries recommended workflows\" align=\"right\" width=75%>\n",
+    "<img src='./assets/240927_multichan.png' alt=\"Multichannel timeseries recommended workflows\" align=\"right\" width=75%>\n",
     "\n",
     "The [recommended multichannel timeseries notebook](./multichan.ipynb) provides a workflow for processing and analyzing multichannel timeseries datasets that fit in memory.\n",
     "\n",

diff --git a/multichannel_timeseries/large_multichan.ipynb b/multichannel_timeseries/large_multichan.ipynb
@@ -129,7 +129,7 @@
     "PYRAMID_FILE = f\"{DATA_PATH.stem}.zarr\"\n",
     "PYRAMID_PATH = DATA_DIR / PYRAMID_FILE\n",
     "print(f'Local Original Data Path: {DATA_PATH}')\n",
-    "print(f'Pyramid Path To Be Created: {PYRAMID_PATH}')"
+    "print(f'Pyramid Path: {PYRAMID_PATH}')"
    ]
   },
   {
@@ -278,15 +278,6 @@
     "</div>"
    ]
   },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "#TODO. Check with Andrew about creating a delayed version of this.. I know it's downsampling in a chunked manner, but I think it then lives entirely in memory prior to getting dumped to disk"
-   ]
-  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -295,10 +286,9 @@
    "source": [
     "%%time\n",
     "\n",
-    "FACTORS = [1, 2, 4, 8, 16, 32, 64, 128, 256]\n",
-    "\n",
-    "# TODO: find better principled way to determine factors.. The following doesn't work as the number of channels scales\n",
-    "# FACTORS = list(np.array([1, 2, 4, 8, 16, 32, 64, 128, 256]) ** (len(ts_ds[\"channel\"]) // 4))\n",
+    "# The FACTORS will depend on the size of your dataset, available disk size, and the resolution needed\n",
+    "# For this demo, we are arbitrarily choosing 8 levels and scaling by quadrupling each preceding factor.\n",
+    "FACTORS = [4**i for i in range(8)]\n",
     "\n",
     "def _help_downsample(data, time, n_out):\n",
     "    \"\"\"\n",
@@ -482,14 +472,13 @@
     "        pyramid_level = num_levels - 1\n",
     "        size = time_da.size\n",
     "    else:\n",
-    "        #TODO: explore a more efficient way to determine resulting size, avoiding\n",
     "        sizes = np.array([\n",
     "            _extract_ds(ts_dt, pyramid_level)[\"time\"].sel(time=time_slice).size\n",
     "            for pyramid_level in range(num_levels)\n",
     "        ])\n",
     "        diffs = sizes - width\n",
     "        pyramid_level = np.argmin(np.where(diffs >= 0, diffs, np.inf)) # nearest higher-resolution level\n",
-    "        # pyramid_level = np.argmin(np.abs(np.array(sizes) - width)) # nearest, regardless of direction\n",
+    "        # pyramid_level = np.argmin(np.abs(np.array(sizes) - width)) # nearest (higher or lower resolution) level\n",
     "        size = sizes[pyramid_level]\n",
     "    \n",
     "    title = (\n",
@@ -644,7 +633,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.11.9"
+   "version": "3.11.10"
   }
  },
  "nbformat": 4,