README.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# LunarLander v2\n",
    "## The description of the task\n",
    "Landing pad is always at coordinates (0,0). Coordinates are the first two numbers in state vector. Reward for moving from the top of the screen to landing pad and zero speed is about 100..140 points. If lander moves away from landing pad it loses reward back. Episode finishes if the lander crashes or comes to rest, receiving additional -100 or +100 points. Each leg ground contact is +10. Firing main engine is -0.3 points each frame. Solved is 200 points. Landing outside landing pad is possible. Fuel is infinite, so an agent can learn to fly and then land on its first attempt. Four discrete actions available: do nothing, fire left orientation engine, fire main engine, fire right orientation engine.\n",
    "\n",
    "Source: https://gym.openai.com/envs/LunarLander-v2/\n",
    "\n",
    "## Approach to resolve the task\n",
    "\n",
    "To resolve the task has been used a deep Q-network (DQN), implemented by Keras framework. The best model and optimal hyper parameters found by a grid search, running `lunarlander.py` script."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![Ideal landing. 2010th episode](lunarlander.gif)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The search grid can be set by editing `lunarlander.py`, for the instance:\n",
    "```python\n",
    "    # Grid parameters\n",
    "    num_learns = [1, 2]\n",
    "    epochs = [1,4]\n",
    "    lookbacks = [10000, 20000, 30000]\n",
    "    batch_sizes = [32]\n",
    "```\n",
    "where:\n",
    "* `num_learn` - how many independent batches and fitting procedure will be used to fit the model after the each step;\n",
    "* `epochs` - the number of epochs for each running the fit procedure;\n",
    "* `lookbacks` - the deep of a data history used to get the batches;\n",
    "* `batch_sizes` - the size of batch to the model;\n",
    "\n",
    "### Running the script\n",
    "Activating a conda virtual environment and running the script:\n",
    "```python\n",
    "source activate /opt/envs/ML\n",
    "setsid python lunarlander.py > ll.log 2>&1\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Checking the perfomance of model(s):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 210: [10  0  0] -78 -48 -58 [0 7 3 0] -2 2 -1\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 410: [0 8 2] 89 215 178 [0 2 1 7] -0 100 100\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 610: [2 6 2] -14 225 143 [1 1 0 8] -100 100 100\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 810: [3 3 4] -63 267 113 [2 1 2 5] -100 100 50\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 1010: [ 0  0 10] 239 285 266 [ 0  0  0 10] 100 100 100\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 1210: [0 2 8] 102 299 250 [0 1 1 8] -0 100 100\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 1410: [ 0  0 10] 237 308 281 [ 0  0  0 10] 100 100 100\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 1610: [ 0  0 10] 217 296 250 [ 0  0  0 10] 100 100 100\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 1810: [ 0  0 10] 230 305 283 [ 0  0  0 10] 100 100 100\n",
      "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 2010: [ 0  0 10] 257 295 271 [ 0  0  0 10] 100 100 100\n"
     ]
    }
   ],
   "source": [
    "!grep TEST ll.log"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "where, for instance `ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32 TEST 410: [0 8 2] 89 215 178 [0 2 1 7] -0 100 100` row:\n",
    "* `128relu-64relu-4linear` - dense neural network with 128 input, 64 hidden and 4 output neurons with `relu` activation for first two layers and `linear` activation for output layer respectively;\n",
    "* `nl1` - paramter `num_learn`=1;\n",
    "* `epochs1` - one epoch for each fitting;\n",
    "* `lb30000` - using 30000 last steps in the data history to get the batch;\n",
    "* `bs32` - batch size is set to 32;\n",
    "* `410` - the test running with 10 episodes and the epsilon parameter equal 0 after 400 episodes of training;\n",
    "* `[0 8 2]` - 0 episodes are finished with summarized reward <= 0, 8 episodes >0 and <= 200 and 2 episodes with summarized reward > 200;\n",
    "* `89` - the minimal summarized reward;\n",
    "* `215` - the maximal summarized reward;\n",
    "* `178` - the median summarized reward;\n",
    "* `[0 2 1 7]` - 0 episodes were finished with the last reward equal -100, -100 < 2 episodes <=0, 0 < 1 episode < 100, 7 episodes with last reward equal 100;\n",
    "* `-0` - a liitle less than 0, the minimal last reward;\n",
    "* `100` - the maximal last reward;\n",
    "* `100` - the median last reward;"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Copying the perspective models by episodes:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "!cp data/ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-410.h5 models/.\n",
    "!cp data/ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1010.h5 models/.\n",
    "!cp data/ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1410.h5 models/.\n",
    "!cp data/ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1610.h5 models/.\n",
    "!cp data/ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1810.h5 models/.\n",
    "!cp data/ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-2010.h5 models/."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[The progress in learning by episodes](https://youtu.be/GSNgYPWdHQA).\n",
    "\n",
    "[![Skills by episodes](https://img.youtube.com/vi/GSNgYPWdHQA/0.jpg)](https://youtu.be/GSNgYPWdHQA/watch?v=GSNgYPWdHQA)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Detailed analysis of the performance of the perspective models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gym\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import glob\n",
    "import ntpath\n",
    "import re\n",
    "\n",
    "from keras.models import load_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "sns.set(rc={'figure.figsize':(11.7,8.27)})\n",
    "sns.set_style(\"whitegrid\", {'grid.linestyle': '--'})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "MODELDIR='models'\n",
    "model_file_mask = f'{MODELDIR}/*h5'\n",
    "\n",
    "df = pd.DataFrame(\n",
    "    columns={\n",
    "        'model': str,\n",
    "        'episode': np.uint,\n",
    "        'step': np.uint,\n",
    "        'reward': np.float,\n",
    "        'done': np.bool\n",
    "    }\n",
    ")\n",
    "\n",
    "env = gym.make('LunarLander-v2')\n",
    "\n",
    "# N is number of running for the each model\n",
    "N=100\n",
    "\n",
    "RE_MODEL_PATTERN = r'll-(\\S+)-nl(\\d+)-epochs(\\d+)-lb(\\d+)-bs(\\d+)-model-(\\d+)'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def parce_model(line):\n",
    "    g = re.match(RE_MODEL_PATTERN, str(line))\n",
    "    if g:\n",
    "        grps = g.groups()\n",
    "        return grps"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run the each model N times:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "for file in glob.glob(model_file_mask):\n",
    "    name = ntpath.basename(file).split('.')[0]\n",
    "    model_par = parce_model(name)\n",
    "    with open(file, encoding='utf-8') as f:\n",
    "        model = load_model(file)\n",
    "        for i in range(N):\n",
    "            s = env.reset()\n",
    "            s = s.reshape(1, 8)\n",
    "            done = False\n",
    "            step = 1\n",
    "            while not done:\n",
    "                s_pred = model.predict(s)\n",
    "                a = np.argmax(s_pred)\n",
    "                new_s, r, done, _ = env.step(a)\n",
    "                s = new_s.reshape(1, 8)\n",
    "                df = df.append(\n",
    "                    {\n",
    "                        'model': name,\n",
    "                        'nn': model_par[0],\n",
    "                        'nl': model_par[1],\n",
    "                        'epochs': model_par[2],\n",
    "                        'lb': model_par[3],\n",
    "                        'bs': model_par[4],\n",
    "                        'train_ep': model_par[5],\n",
    "                        'episode': i,\n",
    "                        'step': step,\n",
    "                        'reward': r,\n",
    "                        'done': done\n",
    "                    },\n",
    "                    ignore_index=True\n",
    "                )\n",
    "                step+=1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Getting the last rewards for the each test and groupping by the models:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = df[df['done']].copy()\n",
    "lr_groups = lr.groupby(['model'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Creating bins/histograms for the last rewards"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "lsr_bins = [-np.inf, -100, 0, 100-np.finfo(np.float16).eps, 100]\n",
    "lr['reward_cat'] = lr_groups['reward'].apply(pd.cut, bins=lsr_bins, labels=['supneg', 'neg', 'pos', 'suppos'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>reward_cat</th>\n",
       "      <th>neg</th>\n",
       "      <th>pos</th>\n",
       "      <th>supneg</th>\n",
       "      <th>suppos</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1010</th>\n",
       "      <td>5.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>90.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1410</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1610</th>\n",
       "      <td>6.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>91.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1810</th>\n",
       "      <td>2.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>98.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-2010</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "reward_cat                                          neg  pos  supneg  suppos\n",
       "model                                                                       \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  5.0  3.0     2.0    90.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  NaN     NaN   100.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  6.0  3.0     NaN    91.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  2.0  NaN     NaN    98.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  NaN     NaN   100.0"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lrhist = lr.groupby('model')['reward_cat'].value_counts().unstack()\n",
    "lrhist.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### The models that only have max possible last rewards"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>reward_cat</th>\n",
       "      <th>neg</th>\n",
       "      <th>pos</th>\n",
       "      <th>supneg</th>\n",
       "      <th>suppos</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1410</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-2010</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "reward_cat                                          neg  pos  supneg  suppos\n",
       "model                                                                       \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  NaN     NaN   100.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  NaN     NaN   100.0"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lrhist[lrhist['suppos'] == N]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Getting the summarized rewards for the each test and grouping by the models:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "sr = pd.DataFrame()\n",
    "sr['sumreward'] = df.groupby(['model', 'episode'])['reward'].sum()\n",
    "sr = sr.reset_index()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Creating bins/histograms for the summarized rewards:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "sr_bins = [-np.inf, 0, 200-np.finfo(np.float).eps, np.inf]\n",
    "sr_groups = sr.groupby(['model'])\n",
    "sr['sr_cat'] = sr_groups['sumreward'].apply(pd.cut, bins=sr_bins, labels=['neg', 'pos', 'suppos'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>sr_cat</th>\n",
       "      <th>neg</th>\n",
       "      <th>pos</th>\n",
       "      <th>suppos</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1010</th>\n",
       "      <td>2.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>89.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1410</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1610</th>\n",
       "      <td>NaN</td>\n",
       "      <td>9.0</td>\n",
       "      <td>91.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1810</th>\n",
       "      <td>NaN</td>\n",
       "      <td>2.0</td>\n",
       "      <td>98.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-2010</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "sr_cat                                              neg  pos  suppos\n",
       "model                                                               \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  2.0  9.0    89.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  NaN   100.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  9.0    91.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  2.0    98.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  NaN   100.0"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "srhist = sr.groupby('model')['sr_cat'].value_counts().unstack()\n",
    "srhist.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### The models that only have the summarized rewards bigger than 200"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>sr_cat</th>\n",
       "      <th>neg</th>\n",
       "      <th>pos</th>\n",
       "      <th>suppos</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1410</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-2010</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "sr_cat                                              neg  pos  suppos\n",
       "model                                                               \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  NaN   100.0\n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  NaN  NaN   100.0"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "srhist[srhist['suppos'] == N]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Visualization of the distribution of summarized rewards for the each models/episodes: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f931ceddd68>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 842.4x595.44 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set(rc={'figure.figsize':(11.7,8.27)})\n",
    "sns.set_style(\"whitegrid\", {'grid.linestyle': '--'})\n",
    "sns.violinplot(x='sumreward', y='model', data=sr, inner='quartile')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Minimal rewards of models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>episode</th>\n",
       "      <th>sumreward</th>\n",
       "      <th>sr_cat</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-2010</th>\n",
       "      <td>0</td>\n",
       "      <td>229.980929</td>\n",
       "      <td>suppos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1410</th>\n",
       "      <td>0</td>\n",
       "      <td>226.773341</td>\n",
       "      <td>suppos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1810</th>\n",
       "      <td>0</td>\n",
       "      <td>159.111153</td>\n",
       "      <td>pos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1610</th>\n",
       "      <td>0</td>\n",
       "      <td>120.745474</td>\n",
       "      <td>pos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-410</th>\n",
       "      <td>0</td>\n",
       "      <td>-87.228161</td>\n",
       "      <td>neg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1010</th>\n",
       "      <td>0</td>\n",
       "      <td>-180.545205</td>\n",
       "      <td>neg</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    episode   sumreward  \\\n",
       "model                                                                     \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...        0  229.980929   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...        0  226.773341   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...        0  159.111153   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...        0  120.745474   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...        0  -87.228161   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...        0 -180.545205   \n",
       "\n",
       "                                                    sr_cat  \n",
       "model                                                       \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  suppos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  suppos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...     pos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...     pos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...     neg  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...     neg  "
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sr.groupby('model').min().sort_values(['sumreward'], ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Maximal rewards of models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>episode</th>\n",
       "      <th>sumreward</th>\n",
       "      <th>sr_cat</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1010</th>\n",
       "      <td>99</td>\n",
       "      <td>317.687348</td>\n",
       "      <td>suppos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1610</th>\n",
       "      <td>99</td>\n",
       "      <td>314.416391</td>\n",
       "      <td>suppos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1410</th>\n",
       "      <td>99</td>\n",
       "      <td>311.419706</td>\n",
       "      <td>suppos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1810</th>\n",
       "      <td>99</td>\n",
       "      <td>303.393806</td>\n",
       "      <td>suppos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-2010</th>\n",
       "      <td>99</td>\n",
       "      <td>300.146494</td>\n",
       "      <td>suppos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-410</th>\n",
       "      <td>99</td>\n",
       "      <td>243.872908</td>\n",
       "      <td>suppos</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    episode   sumreward  \\\n",
       "model                                                                     \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...       99  317.687348   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...       99  314.416391   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...       99  311.419706   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...       99  303.393806   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...       99  300.146494   \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...       99  243.872908   \n",
       "\n",
       "                                                    sr_cat  \n",
       "model                                                       \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  suppos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  suppos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  suppos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  suppos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  suppos  \n",
       "ll-128relu-64relu-4linear-nl1-epochs1-lb30000-b...  suppos  "
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sr.groupby('model').max().sort_values(['sumreward'], ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = load_model('models/ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-1410.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R: 315.0535930887596, r: 100\n"
     ]
    }
   ],
   "source": [
    "env = gym.make('LunarLander-v2')\n",
    "env.mode = 'human'\n",
    "s = env.reset()\n",
    "s = s.reshape(1, 8)\n",
    "#env.render()\n",
    "done = False\n",
    "r_sum = 0\n",
    "while not done:\n",
    "    s_pred = model.predict(s)\n",
    "    a = np.argmax(s_pred)\n",
    "    new_s, r, done, _ = env.step(a)\n",
    "    #env.render()\n",
    "    s = new_s.reshape(1, 8)\n",
    "    r_sum += r\n",
    "    \n",
    "print(f'R: {r_sum}, r: {r}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "env.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = load_model('models/ll-128relu-64relu-4linear-nl1-epochs1-lb30000-bs32-model-2010.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R: 256.658205470644, r: 100\n"
     ]
    }
   ],
   "source": [
    "env = gym.make('LunarLander-v2')\n",
    "# env.mode = 'human'\n",
    "s = env.reset()\n",
    "s = s.reshape(1, 8)\n",
    "env.render()\n",
    "done = False\n",
    "r_sum = 0\n",
    "while not done:\n",
    "    s_pred = model.predict(s)\n",
    "    a = np.argmax(s_pred)\n",
    "    new_s, r, done, _ = env.step(a)\n",
    "    env.render()\n",
    "    s = new_s.reshape(1, 8)\n",
    "    r_sum += r\n",
    "print(f'R: {r_sum}, r: {r}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "env.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusion\n",
    "The agent reached a good level of the performance around 400th episode of the training but sometimes did mistakes. Then, espacialy after the 1010th episode, it looks that the agent was trying to find out optimal balance between maximal performance and a stable result."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}