decoding_from_sweep_response.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## requirements\n",
    "# allensdk\n",
    "# scikit-learn > 0.19\n",
    "# xarray"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from allensdk.core.brain_observatory_cache import BrainObservatoryCache"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "boc = BrainObservatoryCache(manifest_file='/local1/data/boc/manifest.json',)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "oeid = 541206592\n",
    "\n",
    "# Initializations:\n",
    "nwb_dataset = boc.get_ophys_experiment_data(oeid)\n",
    "\n",
    "# # Get Data:\n",
    "# timestamps, dff = nwb_dataset.get_dff_traces()\n",
    "# neuron_ids = nwb_dataset.get_cell_specimen_ids()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import xarray as xr\n",
    "\n",
    "from allensdk.brain_observatory.natural_scenes import NaturalScenes\n",
    "\n",
    "def get_ns_msrx(nwb_dataset):\n",
    "    ns = NaturalScenes(nwb_dataset)\n",
    "    mean_sweep_response = ns.mean_sweep_response.copy()\n",
    "    \n",
    "    # I don't know what dx is. goodbye!\n",
    "    mean_sweep_response.drop('dx',axis=1,inplace=True)\n",
    "    \n",
    "    # annotate the dataframe with useful indices and columns\n",
    "    time = pd.Series(\n",
    "        ns.timestamps[ns.stim_table['start']],\n",
    "        name='time',\n",
    "    )\n",
    "    neurons = pd.Series(\n",
    "        ns.cell_id,\n",
    "        name='neuron',\n",
    "    )\n",
    "    mean_sweep_response.set_index(time,inplace=True)\n",
    "    mean_sweep_response.columns = neurons\n",
    "    \n",
    "    # convert this to xarray & annotate images\n",
    "    srx = xr.DataArray(mean_sweep_response)\n",
    "    srx.coords['natural_image'] = ('time',ns.stim_table['frame'])\n",
    "    \n",
    "    return srx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<xarray.DataArray (time: 5950, neuron: 154)>\n",
      "array([[  2.750398,   3.113332,   3.283231, ...,   3.246232,  17.838305,\n",
      "         44.883263],\n",
      "       [  5.472741,   4.520462,   1.848134, ...,   3.700325,  41.864319,\n",
      "         55.052734],\n",
      "       [  4.938696,   1.872071,   0.822514, ...,   1.446858,  19.081219,\n",
      "         -0.731454],\n",
      "       ..., \n",
      "       [  0.984122,  -1.035176,   2.519683, ...,  -0.894795,  14.030047,\n",
      "          1.291967],\n",
      "       [  1.705522,   1.379635,   1.116845, ...,   0.169055,   6.26528 ,\n",
      "          1.80034 ],\n",
      "       [  0.420941,   2.543081,   2.991403, ...,   0.980473,  -0.720777,\n",
      "          4.779262]])\n",
      "Coordinates:\n",
      "  * time           (time) float64 545.2 545.5 545.7 546.0 546.2 546.5 546.7 ...\n",
      "  * neuron         (neuron) int64 541510267 541510270 541510307 541510405 ...\n",
      "    natural_image  (time) int64 92 27 52 37 103 1 54 19 54 -1 74 115 44 88 ...\n"
     ]
    }
   ],
   "source": [
    "msrx = get_ns_msrx(nwb_dataset)\n",
    "print msrx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "def decode(msrx):\n",
    "    \n",
    "    # get features and output\n",
    "    X = msrx.data\n",
    "    y = msrx['natural_image']\n",
    "    \n",
    "    # split training & testing\n",
    "    X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.2,stratify=y)\n",
    "    \n",
    "    # do the classification\n",
    "    lm = LogisticRegression(\n",
    "        solver='saga',\n",
    "        multi_class='ovr',\n",
    "        penalty='l1',\n",
    "        n_jobs=-1,\n",
    "    )\n",
    "    lm.fit(X_train,y_train)\n",
    "    return lm.score(X_test,y_test) * len(np.unique(y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "19.9\n",
      "1 loop, best of 1: 50.1 s per loop\n"
     ]
    }
   ],
   "source": [
    "%%timeit -n 1 -r 1\n",
    "print decode(msrx)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:jk]",
   "language": "python",
   "name": "conda-env-jk-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}