stronghold_rnn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "2b9ebf0e-0f19-4113-a6d7-e9208c6a8bee",
   "metadata": {},
   "source": [
    "# StrongholdNet: Train an RNN (LSTM) to navigate through a Stronghold\n",
    "\n",
    "The idea is that we interpret the (shortest) path from any room in the stronghold to the portal room as *sequential data* that we feed to an RNN."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "18228814-10ff-4299-a6f5-7a4cbfd606b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras as K\n",
    "#from sklearn.preprocessing import OneHotEncoder\n",
    "#from sklearn.compose import ColumnTransformer\n",
    "from sklearn.model_selection import train_test_split\n",
    "from dataset_rnn import parse_tree_generator, print_stronghold_tree\n",
    "from anytree import Node, RenderTree, Walker\n",
    "from anytree.search import find_by_attr, findall_by_attr\n",
    "import random\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "003064e4-a174-4380-ad89-da53c0ebeb4b",
   "metadata": {},
   "outputs": [],
   "source": [
    "os.environ['CUDA_VISIBLE_DEVICES'] = '-1'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "0bd959bd-bb1f-488e-87cc-51578759aa0b",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('100k_dataset_rnn.csv', delimiter=' ')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "833f9852-8de5-4cbd-a9b5-687b29de613b",
   "metadata": {
    "tags": []
   },
   "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>stronghold</th>\n",
       "      <th>room</th>\n",
       "      <th>entry</th>\n",
       "      <th>orientation</th>\n",
       "      <th>parent_room</th>\n",
       "      <th>parent_exit</th>\n",
       "      <th>child_room_1</th>\n",
       "      <th>child_room_2</th>\n",
       "      <th>child_room_3</th>\n",
       "      <th>child_room_4</th>\n",
       "      <th>child_room_5</th>\n",
       "      <th>exit</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>1</td>\n",
       "      <td>S</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>1</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>1</td>\n",
       "      <td>S</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>2</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>2</td>\n",
       "      <td>W</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>1</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>0</td>\n",
       "      <td>W</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>1</td>\n",
       "      <td>LeftTurn</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>2</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>1</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>1</td>\n",
       "      <td>PortalRoom</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>1</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>1</td>\n",
       "      <td>W</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>2</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>2</td>\n",
       "      <td>N</td>\n",
       "      <td>ChestCorridor</td>\n",
       "      <td>1</td>\n",
       "      <td>PrisonHall</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>ChestCorridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1</td>\n",
       "      <td>ChestCorridor</td>\n",
       "      <td>1</td>\n",
       "      <td>N</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>1</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>1</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>1</td>\n",
       "      <td>W</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>4</td>\n",
       "      <td>ChestCorridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>1</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>4</td>\n",
       "      <td>S</td>\n",
       "      <td>Start</td>\n",
       "      <td>1</td>\n",
       "      <td>PrisonHall</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>None</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>ChestCorridor</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>1</td>\n",
       "      <td>ChestCorridor</td>\n",
       "      <td>0</td>\n",
       "      <td>W</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>5</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>1</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>0</td>\n",
       "      <td>W</td>\n",
       "      <td>ChestCorridor</td>\n",
       "      <td>1</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>SpiralStaircase</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>1</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>0</td>\n",
       "      <td>W</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>1</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>1</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>0</td>\n",
       "      <td>W</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>1</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>None</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>None</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>1</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>3</td>\n",
       "      <td>PortalRoom</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>2</td>\n",
       "      <td>SpiralStaircase</td>\n",
       "      <td>1</td>\n",
       "      <td>W</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>1</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>2</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>1</td>\n",
       "      <td>W</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>2</td>\n",
       "      <td>SpiralStaircase</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>SmallCorridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>2</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>2</td>\n",
       "      <td>N</td>\n",
       "      <td>FiveWayCrossing</td>\n",
       "      <td>5</td>\n",
       "      <td>LeftTurn</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>2</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>0</td>\n",
       "      <td>E</td>\n",
       "      <td>SquareRoom</td>\n",
       "      <td>3</td>\n",
       "      <td>LeftTurn</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>2</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>2</td>\n",
       "      <td>Stairs</td>\n",
       "      <td>SmallCorridor</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>2</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>0</td>\n",
       "      <td>E</td>\n",
       "      <td>Corridor</td>\n",
       "      <td>3</td>\n",
       "      <td>PortalRoom</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>3</td>\n",
       "      <td>ChestCorridor</td>\n",
       "      <td>1</td>\n",
       "      <td>E</td>\n",
       "      <td>RightTurn</td>\n",
       "      <td>1</td>\n",
       "      <td>SpiralStaircase</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    stronghold             room  entry orientation      parent_room  \\\n",
       "0            0        RightTurn      1           S           Stairs   \n",
       "1            0           Stairs      1           S         Corridor   \n",
       "2            0         Corridor      2           W  FiveWayCrossing   \n",
       "3            0         Corridor      0           W         Corridor   \n",
       "4            0         Corridor      0           S         Corridor   \n",
       "5            0         Corridor      0           S         Corridor   \n",
       "6            0        RightTurn      0           S         Corridor   \n",
       "7            1        RightTurn      1           W       SquareRoom   \n",
       "8            1       SquareRoom      2           N    ChestCorridor   \n",
       "9            1    ChestCorridor      1           N        RightTurn   \n",
       "10           1        RightTurn      1           W  FiveWayCrossing   \n",
       "11           1  FiveWayCrossing      4           S            Start   \n",
       "12           1    ChestCorridor      0           W  FiveWayCrossing   \n",
       "13           1       SquareRoom      0           W    ChestCorridor   \n",
       "14           1           Stairs      0           W       SquareRoom   \n",
       "15           1  FiveWayCrossing      0           W           Stairs   \n",
       "16           1       SquareRoom      0           S  FiveWayCrossing   \n",
       "17           2  SpiralStaircase      1           W       SquareRoom   \n",
       "18           2       SquareRoom      1           W       SquareRoom   \n",
       "19           2       SquareRoom      2           N  FiveWayCrossing   \n",
       "20           2         Corridor      0           E       SquareRoom   \n",
       "21           2         Corridor      0           N         Corridor   \n",
       "22           2         Corridor      0           E         Corridor   \n",
       "23           3    ChestCorridor      1           E        RightTurn   \n",
       "\n",
       "    parent_exit     child_room_1     child_room_2   child_room_3 child_room_4  \\\n",
       "0             1           Stairs             None           None         None   \n",
       "1             2        RightTurn             None           None         None   \n",
       "2             1         Corridor           Stairs      RightTurn         None   \n",
       "3             1         LeftTurn         Corridor           None         None   \n",
       "4             2         Corridor             None           None         None   \n",
       "5             1        RightTurn             None           None         None   \n",
       "6             1       PortalRoom             None           None         None   \n",
       "7             2           Stairs             None           None         None   \n",
       "8             1       PrisonHall        RightTurn  ChestCorridor         None   \n",
       "9             1       SquareRoom             None           None         None   \n",
       "10            4    ChestCorridor             None           None         None   \n",
       "11            1       PrisonHall       SquareRoom           None    RightTurn   \n",
       "12            5       SquareRoom             None           None         None   \n",
       "13            1           Stairs  SpiralStaircase      RightTurn         None   \n",
       "14            1  FiveWayCrossing             None           None         None   \n",
       "15            1         Corridor             None     SquareRoom         None   \n",
       "16            3       PortalRoom  FiveWayCrossing       Corridor         None   \n",
       "17            1         Corridor             None           None         None   \n",
       "18            2  SpiralStaircase         Corridor  SmallCorridor         None   \n",
       "19            5         LeftTurn       SquareRoom       Corridor         None   \n",
       "20            3         LeftTurn         Corridor           None         None   \n",
       "21            2           Stairs    SmallCorridor       Corridor         None   \n",
       "22            3       PortalRoom             None           None         None   \n",
       "23            1  SpiralStaircase             None           None         None   \n",
       "\n",
       "     child_room_5  exit  \n",
       "0            None     0  \n",
       "1            None     0  \n",
       "2            None     1  \n",
       "3            None     2  \n",
       "4            None     1  \n",
       "5            None     1  \n",
       "6            None     1  \n",
       "7            None     0  \n",
       "8            None     0  \n",
       "9            None     0  \n",
       "10           None     0  \n",
       "11  ChestCorridor     5  \n",
       "12           None     1  \n",
       "13           None     1  \n",
       "14           None     1  \n",
       "15      RightTurn     3  \n",
       "16           None     1  \n",
       "17           None     0  \n",
       "18           None     0  \n",
       "19           None     3  \n",
       "20           None     2  \n",
       "21           None     3  \n",
       "22           None     1  \n",
       "23           None     0  "
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(24)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "75d9fe84-c17b-4488-856f-b0bde410d311",
   "metadata": {},
   "source": [
    "Let's apply a custom one-hot encoding…"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "e03d166c-80ff-4b2c-9859-871ecbddc579",
   "metadata": {},
   "outputs": [],
   "source": [
    "def onehot(df: pd.DataFrame):\n",
    "    room_to_vector = {\n",
    "            'Corridor': 0,\n",
    "            'PrisonHall': 1,\n",
    "            'LeftTurn': 2,\n",
    "            'RightTurn': 3,\n",
    "            'SquareRoom': 4,\n",
    "            'Stairs': 5,\n",
    "            'SpiralStaircase': 6,\n",
    "            'FiveWayCrossing': 7,\n",
    "            'ChestCorridor': 8,\n",
    "            'Library': 9,\n",
    "            'PortalRoom': 10,\n",
    "            'SmallCorridor': 11,\n",
    "            'Start': 12,\n",
    "            'None': 13}\n",
    "    orientation_to_vector = {\n",
    "            'N': 0,\n",
    "            'S': 1,\n",
    "            'E': 2,\n",
    "            'W': 3}\n",
    "    room_columns = [\n",
    "            'room',\n",
    "            'parent_room',\n",
    "            'child_room_1',\n",
    "            'child_room_2',\n",
    "            'child_room_3',\n",
    "            'child_room_4',\n",
    "            'child_room_5']\n",
    "    exit_columns = [\n",
    "            'entry',\n",
    "            'parent_exit']\n",
    "    orientation_columns = [\n",
    "            'orientation']\n",
    "    n_exits = 6\n",
    "    \n",
    "    df_ = df.copy()\n",
    "\n",
    "    # one-hot rooms\n",
    "    for c in room_columns:\n",
    "        for v in room_to_vector:\n",
    "            df_[c] = df_[c].replace(v, room_to_vector[v])\n",
    "        onehot = pd.DataFrame(\n",
    "                K.utils.to_categorical(df_[c], num_classes=len(room_to_vector)),\n",
    "                columns=[ c + '_is_' + j for j in room_to_vector ])\n",
    "        df_ = pd.concat([ onehot, df_ ], axis=1)\n",
    "\n",
    "    # one-hot exits\n",
    "    for c in exit_columns:\n",
    "        onehot = pd.DataFrame(\n",
    "                K.utils.to_categorical(df_[c], num_classes=n_exits),\n",
    "                columns=[ c + '_is_' + str(j) for j in range(n_exits) ])\n",
    "        df_ = pd.concat([ onehot, df_ ], axis=1)\n",
    "\n",
    "    # one-hot orientations\n",
    "    for c in orientation_columns:\n",
    "        for v in orientation_to_vector:\n",
    "            df_[c] = df_[c].replace(v, orientation_to_vector[v])\n",
    "        onehot = pd.DataFrame(\n",
    "                K.utils.to_categorical(df_[c], num_classes=len(orientation_to_vector)),\n",
    "                columns=[ c + '_is_' + j for j in orientation_to_vector ])\n",
    "        df_ = pd.concat([ onehot, df_ ], axis=1)\n",
    "        \n",
    "    df_.drop([ k for k in room_columns + exit_columns + orientation_columns ], axis=1, inplace=True)\n",
    "    return df_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "48a715d6-908b-4d59-9c2a-7afb2eb8b236",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_onehot = onehot(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "fb1b9c04-d8dc-42b0-bd12-033b8f7301b9",
   "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>orientation_is_N</th>\n",
       "      <th>orientation_is_S</th>\n",
       "      <th>orientation_is_E</th>\n",
       "      <th>orientation_is_W</th>\n",
       "      <th>parent_exit_is_0</th>\n",
       "      <th>parent_exit_is_1</th>\n",
       "      <th>parent_exit_is_2</th>\n",
       "      <th>parent_exit_is_3</th>\n",
       "      <th>parent_exit_is_4</th>\n",
       "      <th>parent_exit_is_5</th>\n",
       "      <th>...</th>\n",
       "      <th>room_is_SpiralStaircase</th>\n",
       "      <th>room_is_FiveWayCrossing</th>\n",
       "      <th>room_is_ChestCorridor</th>\n",
       "      <th>room_is_Library</th>\n",
       "      <th>room_is_PortalRoom</th>\n",
       "      <th>room_is_SmallCorridor</th>\n",
       "      <th>room_is_Start</th>\n",
       "      <th>room_is_None</th>\n",
       "      <th>stronghold</th>\n",
       "      <th>exit</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 116 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   orientation_is_N  orientation_is_S  orientation_is_E  orientation_is_W  \\\n",
       "0               0.0               1.0               0.0               0.0   \n",
       "1               0.0               1.0               0.0               0.0   \n",
       "2               0.0               0.0               0.0               1.0   \n",
       "3               0.0               0.0               0.0               1.0   \n",
       "4               0.0               1.0               0.0               0.0   \n",
       "\n",
       "   parent_exit_is_0  parent_exit_is_1  parent_exit_is_2  parent_exit_is_3  \\\n",
       "0               0.0               1.0               0.0               0.0   \n",
       "1               0.0               0.0               1.0               0.0   \n",
       "2               0.0               1.0               0.0               0.0   \n",
       "3               0.0               1.0               0.0               0.0   \n",
       "4               0.0               0.0               1.0               0.0   \n",
       "\n",
       "   parent_exit_is_4  parent_exit_is_5  ...  room_is_SpiralStaircase  \\\n",
       "0               0.0               0.0  ...                      0.0   \n",
       "1               0.0               0.0  ...                      0.0   \n",
       "2               0.0               0.0  ...                      0.0   \n",
       "3               0.0               0.0  ...                      0.0   \n",
       "4               0.0               0.0  ...                      0.0   \n",
       "\n",
       "   room_is_FiveWayCrossing  room_is_ChestCorridor  room_is_Library  \\\n",
       "0                      0.0                    0.0              0.0   \n",
       "1                      0.0                    0.0              0.0   \n",
       "2                      0.0                    0.0              0.0   \n",
       "3                      0.0                    0.0              0.0   \n",
       "4                      0.0                    0.0              0.0   \n",
       "\n",
       "   room_is_PortalRoom  room_is_SmallCorridor  room_is_Start  room_is_None  \\\n",
       "0                 0.0                    0.0            0.0           0.0   \n",
       "1                 0.0                    0.0            0.0           0.0   \n",
       "2                 0.0                    0.0            0.0           0.0   \n",
       "3                 0.0                    0.0            0.0           0.0   \n",
       "4                 0.0                    0.0            0.0           0.0   \n",
       "\n",
       "   stronghold  exit  \n",
       "0           0     0  \n",
       "1           0     0  \n",
       "2           0     1  \n",
       "3           0     2  \n",
       "4           0     1  \n",
       "\n",
       "[5 rows x 116 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_onehot.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "33214ad7-177c-42d7-b966-ccb58e4b8019",
   "metadata": {},
   "outputs": [],
   "source": [
    "#for x in df_onehot:\n",
    "#    print(x)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e98abc64-3779-482e-a581-73a4fdc2e0bc",
   "metadata": {},
   "source": [
    "We need to group our sequences by stronghold."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "d48565ee-2962-4647-9768-2b5ee2a42f91",
   "metadata": {},
   "outputs": [],
   "source": [
    "def sequencelify(df, min_sequence=3, max_sequence=12):\n",
    "    for _, s in df.groupby('stronghold'):\n",
    "        s = s.drop('stronghold', axis=1).to_numpy()\n",
    "        for window in range(min_sequence, max_sequence + 1):\n",
    "            X_batch = []\n",
    "            y_batch = []\n",
    "            if s.shape[0] < window:\n",
    "                break\n",
    "            for t in range(s.shape[0] - window + 1):\n",
    "                X = s[t:t+window, :-1]\n",
    "                X = X.reshape(window, X.shape[1])\n",
    "                y = np.array(s[t+window-1, -1])\n",
    "                X_batch.append(X)\n",
    "                y_batch.append(y)\n",
    "            X_batch = np.array(X_batch)\n",
    "            y_batch = np.array(y_batch)\n",
    "            X_batch = X_batch.reshape(s.shape[0] - window + 1, window, s.shape[1] - 1)\n",
    "            y_batch = y_batch.reshape(s.shape[0] - window + 1, 1)\n",
    "            y_batch = K.utils.to_categorical(y_batch, num_classes=6)\n",
    "            yield X_batch, y_batch"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3ffc0825-a1a3-4978-98cf-0898f963b594",
   "metadata": {},
   "source": [
    "Train/test split our data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "b6a9b236-39d5-4291-8740-2992ebec1474",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train, df_test = train_test_split(\n",
    "        df_onehot,\n",
    "        test_size=0.1,\n",
    "        random_state=1337,\n",
    "        shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "09140699-749b-4400-b9f5-eb4e7234e501",
   "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>orientation_is_N</th>\n",
       "      <th>orientation_is_S</th>\n",
       "      <th>orientation_is_E</th>\n",
       "      <th>orientation_is_W</th>\n",
       "      <th>parent_exit_is_0</th>\n",
       "      <th>parent_exit_is_1</th>\n",
       "      <th>parent_exit_is_2</th>\n",
       "      <th>parent_exit_is_3</th>\n",
       "      <th>parent_exit_is_4</th>\n",
       "      <th>parent_exit_is_5</th>\n",
       "      <th>...</th>\n",
       "      <th>room_is_SpiralStaircase</th>\n",
       "      <th>room_is_FiveWayCrossing</th>\n",
       "      <th>room_is_ChestCorridor</th>\n",
       "      <th>room_is_Library</th>\n",
       "      <th>room_is_PortalRoom</th>\n",
       "      <th>room_is_SmallCorridor</th>\n",
       "      <th>room_is_Start</th>\n",
       "      <th>room_is_None</th>\n",
       "      <th>stronghold</th>\n",
       "      <th>exit</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 116 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   orientation_is_N  orientation_is_S  orientation_is_E  orientation_is_W  \\\n",
       "0               0.0               1.0               0.0               0.0   \n",
       "1               0.0               1.0               0.0               0.0   \n",
       "2               0.0               0.0               0.0               1.0   \n",
       "3               0.0               0.0               0.0               1.0   \n",
       "4               0.0               1.0               0.0               0.0   \n",
       "\n",
       "   parent_exit_is_0  parent_exit_is_1  parent_exit_is_2  parent_exit_is_3  \\\n",
       "0               0.0               1.0               0.0               0.0   \n",
       "1               0.0               0.0               1.0               0.0   \n",
       "2               0.0               1.0               0.0               0.0   \n",
       "3               0.0               1.0               0.0               0.0   \n",
       "4               0.0               0.0               1.0               0.0   \n",
       "\n",
       "   parent_exit_is_4  parent_exit_is_5  ...  room_is_SpiralStaircase  \\\n",
       "0               0.0               0.0  ...                      0.0   \n",
       "1               0.0               0.0  ...                      0.0   \n",
       "2               0.0               0.0  ...                      0.0   \n",
       "3               0.0               0.0  ...                      0.0   \n",
       "4               0.0               0.0  ...                      0.0   \n",
       "\n",
       "   room_is_FiveWayCrossing  room_is_ChestCorridor  room_is_Library  \\\n",
       "0                      0.0                    0.0              0.0   \n",
       "1                      0.0                    0.0              0.0   \n",
       "2                      0.0                    0.0              0.0   \n",
       "3                      0.0                    0.0              0.0   \n",
       "4                      0.0                    0.0              0.0   \n",
       "\n",
       "   room_is_PortalRoom  room_is_SmallCorridor  room_is_Start  room_is_None  \\\n",
       "0                 0.0                    0.0            0.0           0.0   \n",
       "1                 0.0                    0.0            0.0           0.0   \n",
       "2                 0.0                    0.0            0.0           0.0   \n",
       "3                 0.0                    0.0            0.0           0.0   \n",
       "4                 0.0                    0.0            0.0           0.0   \n",
       "\n",
       "   stronghold  exit  \n",
       "0           0     0  \n",
       "1           0     0  \n",
       "2           0     1  \n",
       "3           0     2  \n",
       "4           0     1  \n",
       "\n",
       "[5 rows x 116 columns]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "0be2dc3b-8552-4a83-92e9-db8c76361abf",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "df_train: (633893, 116)\n",
      "df_test: (70433, 116)\n"
     ]
    }
   ],
   "source": [
    "print(\"df_train:\", df_train.shape)\n",
    "print(\"df_test:\", df_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "1c1b75e2-b183-433f-a03a-101177cf1a65",
   "metadata": {},
   "outputs": [],
   "source": [
    "n_features = df_train.shape[1] - 2"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8bbe043a-815f-4da4-8c0e-797aef2269fe",
   "metadata": {},
   "source": [
    "Load nav dataset:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "78d518a8-b7f9-4372-ada6-93a81e638408",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_nav = pd.read_csv('100k_dataset_nav.csv')\n",
    "df_nav_onehot = onehot(df_nav)\n",
    "df_nav_train, df_nav_test = train_test_split(\n",
    "        df_nav_onehot,\n",
    "        test_size=0.1,\n",
    "        random_state=1337,\n",
    "        shuffle=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e03185c-807b-4851-90bb-6856b4462521",
   "metadata": {},
   "source": [
    "Build and compile the actual model: an LSTM followed by a dense layer with softmax activation for our 6 exit classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "1a6eda45-170c-405f-af58-a63d1d0700cd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm (LSTM)                  (None, None, 64)          45824     \n",
      "_________________________________________________________________\n",
      "lstm_1 (LSTM)                (None, 64)                33024     \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 6)                 390       \n",
      "=================================================================\n",
      "Total params: 79,238\n",
      "Trainable params: 79,238\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = K.Sequential()\n",
    "#model.add(K.layers.LSTM(\n",
    "#        128,\n",
    "#        batch_input_shape=(None, None, n_features),\n",
    "#        return_sequences=False,\n",
    "#        stateful=False))\n",
    "model.add(K.layers.LSTM(\n",
    "        64,\n",
    "        batch_input_shape=(None, None, n_features),\n",
    "        return_sequences=True,\n",
    "        stateful=False))\n",
    "model.add(K.layers.LSTM(\n",
    "        64,\n",
    "        return_sequences=False,\n",
    "        stateful=False))\n",
    "model.add(K.layers.Dense(\n",
    "        6,\n",
    "        activation='softmax'))\n",
    "model.compile(\n",
    "        optimizer='adam',\n",
    "        loss='categorical_crossentropy',\n",
    "        metrics=['accuracy'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "91078fee-01a6-4f26-81fb-3d17f29cef86",
   "metadata": {},
   "source": [
    "Fit the model…"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "816a2a3e-8a93-4ac3-9860-417053f43a23",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100000/100000 [==============================] - 301s 3ms/step - loss: 0.2486 - accuracy: 0.8811\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x7f261870b700>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(\n",
    "        sequencelify(df_train),\n",
    "        epochs=1,\n",
    "        steps_per_epoch=100000)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e6a65b8d-9c36-42bc-be9f-26a8841e024b",
   "metadata": {},
   "source": [
    "Evaluate the model on test data…"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "5c970648-30c0-4e7b-b75f-a26ed55b8092",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "49696/49696 [==============================] - 48s 962us/step - loss: 0.2446 - accuracy: 0.8865\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.2445562183856964, 0.8864912986755371]"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(sequencelify(df_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f2070cb1-e19e-46d0-86bd-4ac2118bad7e",
   "metadata": {},
   "source": [
    "Save/load model weights."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "6fc5053d-6bc8-4aaa-a0bc-00a2778ffa21",
   "metadata": {},
   "outputs": [],
   "source": [
    "#model.save(\"rnn_11.keras\")\n",
    "model = K.models.load_model(\"rnn_10.keras\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "04b49356-2f3d-45a8-8ce0-476e43b8b841",
   "metadata": {},
   "source": [
    "Let the model navigate new strongholds and learn from its mistakes:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "d7124f74-cb4c-4b03-a225-3feb664ee8ab",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "49696/49696 [==============================] - 50s 1ms/step - loss: 1.4340 - accuracy: 0.2796\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[1.4340380430221558, 0.27957871556282043]"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(sequencelify(df_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "74d48fe5-b741-4a69-8b7c-8ca3260f4cbb",
   "metadata": {},
   "outputs": [],
   "source": [
    "def navigate_and_train(model, root, stronghold, max_visited=50):\n",
    "    X = []\n",
    "    y = []\n",
    "    n_visited = 0\n",
    "    \n",
    "    # locate portal room\n",
    "    portal = find_by_attr(root, 'PortalRoom')\n",
    "    \n",
    "    # skip starter\n",
    "    next_exit = 1\n",
    "    entry = 0\n",
    "    next_room = root.children[0]\n",
    "    \n",
    "    while True:\n",
    "        X.append((\n",
    "                next_room.name,\n",
    "                entry,\n",
    "                next_room.orientation,\n",
    "                next_room.parent.name,\n",
    "                next_room.exit,\n",
    "                *([ c.name for c in next_room.children ] + [ 'None' ] * (5 - len(next_room.children)))))\n",
    "        \n",
    "        # find correct portal label\n",
    "        (upwards, common, downwards) = Walker().walk(next_room, portal)\n",
    "        label = 0 if len(upwards) > 0 else downwards[0].exit\n",
    "        y.append(label)\n",
    "        \n",
    "        # stop if portal room found\n",
    "        if len([r for r in next_room.children if r.name == 'PortalRoom']) > 0:\n",
    "            break\n",
    "            \n",
    "        # stop if we reached max visited rooms\n",
    "        n_visited += 1\n",
    "        if n_visited == max_visited:\n",
    "            break\n",
    "        \n",
    "        # predict next exit\n",
    "        next_exit = predict(model, X)\n",
    "        \n",
    "        # update stuff\n",
    "        entry = 0 if next_exit > 0 else next_room.exit\n",
    "        \n",
    "        # stop if predicting a bad exit\n",
    "        if next_exit > len(next_room.children):\n",
    "            break\n",
    "        \n",
    "        next_room = ([ next_room.parent, *next_room.children ])[next_exit]\n",
    "        \n",
    "        # stop if predicting a bad room\n",
    "        if next_room.name in [ 'Start', 'SmallCorridor', 'Library', 'None' ]:\n",
    "            break\n",
    "            \n",
    "    df = pd.DataFrame(X, columns=(\n",
    "            'room',\n",
    "            'entry',\n",
    "            'orientation',\n",
    "            'parent_room',\n",
    "            'parent_exit',\n",
    "            'child_room_1',\n",
    "            'child_room_2',\n",
    "            'child_room_3',\n",
    "            'child_room_4',\n",
    "            'child_room_5'))\n",
    "    df['stronghold'] = stronghold\n",
    "    df['exit'] = pd.Series(y)\n",
    "    \n",
    "    # append output to csv file\n",
    "    df.to_csv('100k_dataset_nav.csv', mode='a', header=False, index=False)\n",
    "    \n",
    "    df = onehot(df)\n",
    "    model.fit(\n",
    "            sequencelify(df, min_sequence=1, max_sequence=50),\n",
    "            epochs=1,\n",
    "            steps_per_epoch=df.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "40b7924f-5aa8-4e8a-8f1e-7023141629ef",
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(model, X):\n",
    "    df = pd.DataFrame(X, columns=(\n",
    "            'room',\n",
    "            'entry',\n",
    "            'orientation',\n",
    "            'parent_room',\n",
    "            'parent_exit',\n",
    "            'child_room_1',\n",
    "            'child_room_2',\n",
    "            'child_room_3',\n",
    "            'child_room_4',\n",
    "            'child_room_5'))\n",
    "    df = onehot(df)\n",
    "    y_hat = model.predict(df.to_numpy().reshape(1, len(X), n_features))\n",
    "    return np.argmax(y_hat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1e75b585-24e8-49ab-be9a-6bc943d0bd78",
   "metadata": {
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "gen = parse_tree_generator('100k_strongholds_test.txt')\n",
    "stronghold = 0\n",
    "for root in gen:\n",
    "    navigate_and_train(model, root, stronghold)\n",
    "    stronghold += 1"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "99497be7-ab32-4de7-a720-d9af3424efe4",
   "metadata": {},
   "source": [
    "Let's evaluate it again…"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a87df510-3983-450a-8931-3fb4a637f035",
   "metadata": {},
   "outputs": [],
   "source": [
    "model.evaluate(sequencelify(df_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d8f041e6-6cb6-437a-aedc-04072de30496",
   "metadata": {},
   "outputs": [],
   "source": [
    "#model.save(\"rnn_10.keras\")\n",
    "model = K.models.load_model(\"rnn_10.keras\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e16d59c7-a488-4f20-982d-ea857010d2b6",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.9.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}