190314_3layer-trans-inv-sigmoid-reg.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Version SIGMOID-REG: \n",
    "- 44,100 Hz sample rate \n",
    "- regularization with beta = 0.001\n",
    "- 300,000 iterations\n",
    "- 0.95 learning rate decay\n",
    "- sigmoid output function\n",
    "- add a BIAS to output layer\n",
    "- learning rate 0.005\n",
    "\n",
    "**Problem:** regularization too strong with 0.01\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 0. Import the packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os,sys,shutil,mmap\n",
    "import pickle as pickle\n",
    "import numpy as np                                       # fast vectors and matrices\n",
    "import matplotlib.pyplot as plt                          # plotting\n",
    "from scipy.fftpack import fft\n",
    "\n",
    "import time\n",
    "\n",
    "sys.path.insert(0,'lib/')\n",
    "import config\n",
    "\n",
    "from sklearn.metrics import average_precision_score\n",
    "\n",
    "import tensorflow as tf\n",
    "\n",
    "import mir_eval\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "notebook_reference = str(190314_3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "debugging = False\n",
    "delete_ids = [2560, 1792, 2562, 2506, 2564, 2566, 1793, 2568, 2570, 2571, 2572, 2573, 2575, 2576, 2302, 2581, 2582, \n",
    "              2586, 2075, 2076, 2077, 2590, 2079, 2080, 2081, 2594, 2595, 2596, 2603, 2678, 2397, 2608, 2611, 2614, \n",
    "              2104, 2105, 2619, 2620, 2621, 2622, 2112, 2113, 2114, 2627, 2116, 2629, 2118, 2119, 2632, 2633, 2127, \n",
    "              2131, 2138, 2319, 2140, 2405, 2659, 2148, 2149, 2150, 2151, 2154, 2155, 2156, 2157, 2158, 2159, 2160, \n",
    "              2161, 2677, 2166, 2167, 2168, 2169, 2239, 2325, 2177, 2178, 2179, 2180, 1729, 2186, 2242, 2194, 2195, \n",
    "              2196, 1829, 2198, 2200, 2201, 2202, 2203, 2204, 2330, 2207, 2208, 2209, 2210, 2211, 2212, 2213, 2214, \n",
    "              2215, 2588, 2218, 2219, 2220, 2221, 2222, 2224, 2225, 2227, 2228, 2229, 2230, 2231, 2232, 2234, 2591, \n",
    "              2237, 2238, 1727, 1728, 2241, 1730, 2243, 2244, 1733, 1734, 2593, 2248, 1739, 2240, 2082, 1742, 2083, \n",
    "              1749, 1750, 1751, 1752, 1755, 1756, 1757, 1758, 1760, 1763, 1764, 1765, 1766, 1768, 2282, 1771, 1772, \n",
    "              1773, 1775, 1776, 1777, 2292, 2293, 2294, 2295, 2296, 2297, 1919, 1788, 1789, 1790, 2304, 2305, 2307, \n",
    "              2308, 2310, 2313, 2314, 2315, 1805, 2318, 1807, 2320, 2322, 1811, 1812, 1813, 1817, 1818, 2334, 2335, \n",
    "              2336, 1828, 2341, 2342, 2343, 2567, 2345, 2346, 1835, 2348, 2350, 2491, 2357, 2358, 2359, 2364, 2365, \n",
    "              2366, 2368, 2371, 2372, 2373, 2374, 2529, 2376, 2377, 2379, 2443, 2381, 2383, 2384, 1873, 2388, 2389, \n",
    "              2390, 2391, 2392, 2393, 2618, 2398, 2403, 2404, 1893, 2406, 2410, 2411, 2415, 2417, 2147, 2451, 2420, \n",
    "              2422, 2423, 2424, 1916, 2538, 1918, 2431, 2432, 2433, 1922, 1923, 2436, 2441, 2442, 1931, 1932, 2626, \n",
    "              1735, 2285, 1859, 2247, 2117, 2288, 2466, 2444, 2289, 2472, 2473, 2476, 2477, 2478, 2480, 2481, 2482, \n",
    "              2483, 2462, 2486, 2487, 2488, 2490, 2463, 2492, 2494, 2555, 2497, 1822, 2501, 2502, 2504, 2505, 2078, 2507]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Import the data\n",
    "Import the recording and labels of the 331 recordings from the data base.\n",
    "\n",
    "#### 1.1 Import the labels of the data set.\n",
    "The 331 labels are saved in a dictionary whose\n",
    "- keys are integers (the ids of the recordings)\n",
    "- values are intervaltrees (for the different intervals of the recording it gives the played notes between 1 and 128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(config.labels_path, 'rb') as f:\n",
    "    labels = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "331 labels\n"
     ]
    }
   ],
   "source": [
    "print(len(labels), 'labels')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2239\n",
      "<class 'intervaltree.intervaltree.IntervalTree'>\n"
     ]
    }
   ],
   "source": [
    "for key, value in labels.items() :\n",
    "    if key == 2239:\n",
    "        # id of the recording\n",
    "        print (key)\n",
    "        # intervaltree containing for each note intervals in which it is played\n",
    "        print (type(value))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1.2 Import recordings\n",
    "\n",
    "Create a dictionary containing the data (recordings). Use Memory-mapped file objects.\n",
    "For each recording data[rec_id] is a tuple:\n",
    "1. data[rec_id][0] is the memory-mapped file object to the recording. It can be accesed by \n",
    "~~~~~~\n",
    "np.frombuffer(data[rec_id][0], dtype=np.float32).copy()\n",
    "~~~~~~\n",
    "2. data[rec_id][1] is its length (float)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "sz_float = 4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = dict()\n",
    "for record in os.listdir(config.records_path):\n",
    "    fd = os.open(config.records_path + record, os.O_RDONLY)\n",
    "    buff = mmap.mmap(fd, 0, mmap.MAP_SHARED, mmap.PROT_READ)\n",
    "    data[int(record[:-4])] = (buff, len(buff)/sz_float)\n",
    "    os.close(fd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Key: 2494 , Format: <class 'int'>\n",
      "Value: (<mmap.mmap object at 0x7f7f897e9270>, 13100544.0)\n",
      "1. Entry: Memory-mapped file objects , <mmap.mmap object at 0x7f7f897e4618>\n",
      "For example:\n",
      "[ 0.00036621  0.00039673  0.00036621  0.00036621  0.00033569  0.00033569\n",
      "  0.00036621  0.00036621  0.00036621  0.00033569  0.00036621  0.00039673]\n",
      "2. Entry: length of recording 5293440.0 , Type: <class 'float'>\n"
     ]
    }
   ],
   "source": [
    "for key, value in data.items() :\n",
    "    if key == 2494:\n",
    "        # id of the recording\n",
    "        print ('Key:',key,', Format:',type(key))\n",
    "        # its value is a tuple\n",
    "        print ('Value:',value)\n",
    "        # 1. value: reference to recording\n",
    "        print (\"1. Entry: Memory-mapped file objects ,\", data[2239][0])\n",
    "        # to access content use:\n",
    "        x = np.frombuffer(data[2239][0], dtype=np.float32).copy()\n",
    "        print('For example:')\n",
    "        print(x[3005:3017])\n",
    "        print (\"2. Entry: length of recording\",str(data[2239][1]),\", Type:\",type(data[2239][1]))\n",
    "        #print(len(x))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1.3 Split up in the test and train set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are 11 samples in the test set\n",
      "There are 320 samples in the train set\n"
     ]
    }
   ],
   "source": [
    "extended_test_set = True\n",
    "use_mirex = True\n",
    "    \n",
    "# use the small or big test set\n",
    "if extended_test_set:\n",
    "    test_ids = config.test_ids_ext\n",
    "else:\n",
    "    test_ids = config.test_ids\n",
    "\n",
    "# include the mirex dev set in test results\n",
    "if use_mirex:\n",
    "    test_ids = test_ids + config.mirex_id\n",
    "\n",
    "train_ids = [rec_id for rec_id in labels.keys() if rec_id not in test_ids]\n",
    "\n",
    "if debugging == True:\n",
    "    train_ids = [rec_id for rec_id in train_ids if rec_id not in delete_ids]\n",
    "\n",
    "\n",
    "print('There are {} samples in the test set'.format(len(test_ids)))\n",
    "print('There are {} samples in the train set'.format(len(train_ids)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Preprocessing\n",
    "\n",
    "The input data (recordings) are normalized and the possible note range is defines to 128 notes. The lowest note is ..., the highest ... ."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Normalize the recordings\n",
    "normalize = True\n",
    "# Possible notes are from range [base_note, base_note + m)\n",
    "base_note = 0\n",
    "mm = 128"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.1 Data augmentation\n",
    "\n",
    "Always make two small changes to input data by randomly stretching or shrinking our input audio that are not noticable by human hearing, but augment the data set:\n",
    "1. Pitch-shift in the frequency domain: Randomly  shifting  each data points in a minibatch by an integral number of semitones. This reinforces the architectural structure of the translation-invariant network.\n",
    "2. Jittering: continuous shift to each data point. This makes the models more robust to tuning variation between recordings --> regularization\n",
    "\n",
    "Define a function for extracting a sequence and its label as well as applying the data augmentation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# pitching: stretching or shrinking our input audio with linear interpolation\n",
    "pitch_transforms=5 # =/- 5 semitones\n",
    "jitter=.1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pitch-shifting corresponds to a multiplication by $2^{1/12}$ (an octave is equal to 12 semitones and corresponds to a factor 2 in the frequency domain) for each semirone.\n",
    "\n",
    "A scaling factor in the frequency domain implies also a scaling factor in the time domain. (speeding up the recording $\\to$ higher frequency $\\to$ sounds higher.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# function to obtain one data sample from recordings\n",
    "# input:\n",
    "# - rec_id: the id of the recording\n",
    "# - s: frame-number to extract\n",
    "# - window: windowsize\n",
    "# - pitchshift: shift the label by pitchshift halftones\n",
    "# - scaling_factor: scaling factor in frequency domain (includes corrections because of pitch shift and jitter)\n",
    "# - normalize: boolean\n",
    "# output\n",
    "# - x: audio segment of length window\n",
    "# - y: labels, 0-1-vector of length mm\n",
    "def get_data(rec_id,s,window,pitch_shift,scaling_factor, normalize=True):\n",
    "   \n",
    "    x = np.frombuffer(data[rec_id][0][s*sz_float:int(\n",
    "        s+scaling_factor*window)*sz_float], dtype=np.float32).copy()\n",
    "    \n",
    "    if normalize: \n",
    "        x /= np.linalg.norm(x) + config.epsilon\n",
    "\n",
    "    xp = np.arange(window,dtype=np.float32)\n",
    "    # interpolate the discrete function (np.arange(len(x)), x) at the points (scaling_factor*xp)\n",
    "    x = np.interp(scaling_factor*xp,np.arange(len(x),dtype=np.float32),x).astype(np.float32)\n",
    "    # real-valued audio frame with values in the range [−1,1], sampled at 44.1 kHz\n",
    "    # -> x has length window\n",
    "    \n",
    "    y = np.zeros(mm)\n",
    "    # select center of frame\n",
    "    for label in labels[rec_id][s+scaling_factor*window/2]:\n",
    "        y[label.data[1]+pitch_shift-base_note] = 1 # apply pitch shift here (in frequency domain)\n",
    "    \n",
    "    return x,y"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.1 Batch-selection\n",
    "Create the training batch applying the above explained methods for Data augmentation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# function to create an input batch for one training step\n",
    "# input:\n",
    "# - model_stats: model_results-class containing training_ids\n",
    "# - batch_size (integer)\n",
    "# - window: window-length\n",
    "# - pitch_transforms: integer in [-5,5], number of maximum semitones shifted\n",
    "# - jitter: float in [-0.1,0.1], tuning variation\n",
    "# output:\n",
    "# - xmb: batch for a training step, size: (batch_size,1,window,1)\n",
    "# - ymb: corresponding labels to the batch xmb, size: (batch_size,1,mm)\n",
    "def get_training_batch(model_stats, batch_size, window, pitch_transforms=0, jitter=0):\n",
    "    # arrays for input data (batch)\n",
    "    xmbatch = np.empty([batch_size,1,window,1],dtype=np.float32)\n",
    "    ymbatch = np.empty([batch_size,1,mm],dtype=np.float32)\n",
    "        \n",
    "    # picking randomly batch_size recordings\n",
    "    rec_ids = [model_stats.train_ids[r] for r in \\\n",
    "               np.random.randint(0,len(model_stats.train_ids),batch_size)]\n",
    "        \n",
    "    if pitch_transforms > 0:\n",
    "        transform_type = np.random.randint(-pitch_transforms,\n",
    "                                           pitch_transforms,batch_size)\n",
    "\n",
    "    if jitter > 0:\n",
    "        jitter_amount = np.random.uniform(-jitter,jitter,batch_size)\n",
    "\n",
    "    for i in range(batch_size):\n",
    "        scaling_factor = 1\n",
    "        shift = 0\n",
    "        if pitch_transforms > 0:\n",
    "            shift = transform_type[i]\n",
    "            # half-step has a frequency ratio of 2^(1/12)\n",
    "            # (shift + jitter_amount) halfsteps are shifted\n",
    "            if jitter > 0:\n",
    "                scaling_factor = (2.**((shift+jitter_amount[i])/12.))\n",
    "            else:\n",
    "                # only shift the halftones from pitch_transform\n",
    "                scaling_factor = (2.**(shift/12.))\n",
    "        \n",
    "        # multiply in frequency domain by scaling_factor\n",
    "        \n",
    "        # pick a frame for training randomly\n",
    "        s = np.random.randint(0,data[rec_ids[i]][1]-scaling_factor*window)\n",
    "        xmbatch[i,0,:,0],ymbatch[i,0,:] = get_data(rec_ids[i],s,window,pitch_shift=shift,scaling_factor=scaling_factor)\n",
    "        \n",
    "    return xmbatch, ymbatch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2.3 Creation of Training and Test set for protocolling\n",
    "\n",
    "Routines that create sample for the protcoling:\n",
    "1. Routine to obtain the training samples from numpy-file.\n",
    "2. Routine to obtain test samples given the recording IDs of the test samples."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# function to obtain training sample which is already saved in \n",
    "# numpy fromat in file. only normalization is necessary.\n",
    "# input:\n",
    "# - d: window size (integer)\n",
    "# - normalize: boolean\n",
    "# output:\n",
    "# - Xtrain: training sample of size (15000, 1, d, 1)\n",
    "# - Ytrain: corresponding labels of size (15000, 128)\n",
    "def get_training_sample(d=16384, normalize=True):\n",
    "    cachex = config.tmp + 'Xtrain_ext{}.npy'.format(d)\n",
    "    cachey = config.tmp + 'Ytrain_ext{}.npy'.format(d)\n",
    "    #shape Xtrain: (15000, 1, 16384, 1)\n",
    "    #shape Ytrain: (15000, 128)\n",
    "    if os.path.exists(cachex) and os.path.exists(cachey):\n",
    "        Xtrain = np.load(cachex)\n",
    "        Ytrain = np.load(cachey)\n",
    "    else:\n",
    "        print(cachex)\n",
    "        print(cachey)\n",
    "        raise Exception('Unimplemented')\n",
    "\n",
    "    if normalize:\n",
    "        Xtrain = Xtrain.reshape(len(Xtrain),1,d,1)\n",
    "        for i in range(len(Xtrain)):\n",
    "            Xtrain[i] /= np.linalg.norm(Xtrain[i]) + config.epsilon\n",
    "\n",
    "    return Xtrain,Ytrain\n",
    "\n",
    "\n",
    "# function to obtain test sample using get_data(). every sample is split \n",
    "# up in count sequences.\n",
    "# input:\n",
    "# - rec_ids: ids of test samples\n",
    "# - window: size of the input frame\n",
    "# - count: number of sequences to be created for each sample\n",
    "# - fixed_stride: -1 (stride is adapted to sample) or nonnegative integer\n",
    "# - pitch_shift: value by which recording is stretched or shrinked\n",
    "# output:\n",
    "# - Xtest: test sample of size (count*len(rec_ids),1,window,1)\n",
    "# - Ytest: corresponding labels of size (count*len(rec_ids),mm)\n",
    "def get_test_sample(rec_ids, window, count, fixed_stride=-1, pitch_shift=0):\n",
    "    Xtest = np.zeros([count*len(rec_ids),1,window,1])\n",
    "    Ytest = np.zeros([count*len(rec_ids),mm])\n",
    "    offset = 44100\n",
    "    # calculate scaling factor\n",
    "    sf = 2.**(pitch_shift/12.)\n",
    "    # for every sample create 1000 sequences\n",
    "    for i in range(len(rec_ids)):\n",
    "        X = np.zeros([count,1,window,1])\n",
    "        Y = np.zeros([count,mm])\n",
    "        # calculate teh stride in case it is not fixed\n",
    "        stride = (data[rec_ids[i]][1]-offset-int(sf*window))/count if fixed_stride==-1 else fixed_stride\n",
    "        for j in range(count):\n",
    "            X[j,0,:,0],Y[j] = get_data(rec_ids[i],int(offset+j*stride),window,pitch_shift=int(round(pitch_shift)),scaling_factor=sf)\n",
    "        \n",
    "        Xtest[i*count:(i+1)*count] = X\n",
    "        Ytest[i*count:(i+1)*count] = Y\n",
    "    return Xtest, Ytest"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Save the results, losses, precicions and weights at each protocol point inside a class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "class model_results():\n",
    "    def __init__(self, test_ids, train_ids):\n",
    "        # weights\n",
    "        self.weights = dict()\n",
    "        # step\n",
    "        self.iter = 0\n",
    "        # result statistics\n",
    "        self.stats = dict()\n",
    "        self.stats['iter'] = [False,'{:<8}',[]]\n",
    "        # time that has passed since last record\n",
    "        self.stats['time'] = [True,'{:<8.0f}',[]]\n",
    "        self.stats['lr'] = [False,'{:<8.6f}',[]]\n",
    "        # mean squared errors\n",
    "        self.stats['mse_train_1'] = [True,'{:<16.6f}',[]]\n",
    "        self.stats['mse_train'] = [True,'{:<16.6f}',[]]\n",
    "        self.stats['mse_test'] = [True,'{:<16.6f}',[]]\n",
    "        # precicions\n",
    "        self.stats['avp_train'] = [True,'{:<16.6f}',[]]\n",
    "        self.stats['avp_test'] = [True,'{:<16.6f}',[]]\n",
    "        \n",
    "        self.averages = []\n",
    "        \n",
    "        self.test_ids = test_ids\n",
    "        self.train_ids = train_ids\n",
    "\n",
    "    # Register the weights inside the list w_list. \n",
    "    # The list name_list contains the corresponding names of the wariables in w_list\n",
    "    def register_weights(self, w_list, name_list):\n",
    "        if len(w_list) != len(name_list):\n",
    "            print(\"error: input lists need to have same length\")\n",
    "            \n",
    "        for ii in range(len(w_list)):\n",
    "            w = w_list[ii]\n",
    "            name = name_list[ii]\n",
    "            # register the norm of the weights\n",
    "            self.stats['n'+name] = [False,'{:<8.3f}',[]] \n",
    "            # register the weights\n",
    "            self.weights[name] = w\n",
    "            \n",
    "        return None\n",
    "        \n",
    "model_stats = model_results(test_ids, train_ids)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Error measures\n",
    "Define the error measures for comparing the correct labeling Y with the extimated labeling Y_hat.\n",
    "Both labelings are of size (mm) for each segment. The errors are calculated at each protocol point for the training and the test set.\n",
    "1. Mean Squared Error: When applying the model to a sequence or sequences the mean squared error is calculated directly by applying the tensorflow function.\n",
    "~~~~~~~~~~~~~~~~~~~~~\n",
    "tf.reduce_mean(tf.nn.l2_loss( Y_hat - Y ))\n",
    "~~~~~~~~~~~~~~~~~~~~~\n",
    "2. Average Precision: The average precision is calculated with the metric function from the sklear package.\n",
    "~~~~~~~~~~~~~~~~~~~~~\n",
    "sklearn.metrics.average_precision_score\n",
    "~~~~~~~~~~~~~~~~~~~~~\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5. Define the graph\n",
    "\n",
    "Define batch size (batch_size segments at each training step). Also define a small factor wscale to scale the weights. Regularization can be avoided."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size=150\n",
    "if debugging == True:\n",
    "    batch_size = 4\n",
    "out = 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5.1 Define layers"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### First layer: \n",
    "\n",
    "Lenth of the input (segment) is window=16384. Compute a strided convolution with a d=4096-sample recptive field and a stride=512 sample stride. Since every segment has 16384 samples there are 25 regions for each frame.\n",
    "\n",
    "After the convolution for each region a filterbank computation is calculated. \n",
    "As a filter a log=spaced filterbank is chosen (k=512 sine and cosine filters with logarithmically spaced frequencies from 50 hz to 6000 hz) which is going to be applied to all the regions of the convolution of size d=4096"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# variables for input layer\n",
    "\n",
    "# input size of segment, predict notes played at midpoint of segment\n",
    "window=16384\n",
    "# size of regions that outputs the convolution from layer one (receptive field)\n",
    "dd = 4096\n",
    "#stride for first layer convolution\n",
    "stride=512\n",
    "# --> 25 regions\n",
    "\n",
    "# variables for filterbank (= first layer)\n",
    "# number of nodes in first layer (by filterbank get from dd nodes to kk nodes)\n",
    "kk = 512 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_filters(d,k):\n",
    "    # create k filters\n",
    "    # d is the length of the receptive field (input)\n",
    "    x = np.linspace(0, 2*np.pi, d, endpoint=False)\n",
    "    wsin = np.empty((1,d,1,k), dtype=np.float32)\n",
    "    wcos = np.empty((1,d,1,k), dtype=np.float32)\n",
    "    start_freq = 50.\n",
    "    end_freq = 6000.\n",
    "    num_cycles = start_freq*d/44100.                                          \n",
    "    scaling_ind = np.log(end_freq/start_freq)/k\n",
    "    window_mask = 1.0-1.0*np.cos(x)\n",
    "    for ind in range(k):\n",
    "        wsin[0,:,0,ind] = window_mask*np.sin(np.exp(ind*scaling_ind)*num_cycles*x)\n",
    "        wcos[0,:,0,ind] = window_mask*np.cos(np.exp(ind*scaling_ind)*num_cycles*x)\n",
    "            \n",
    "    return wsin,wcos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Second layer:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# variables for second layer\n",
    "d2_x = 1          # lvl2 input dims_x\n",
    "d2_y = 128          # lvl2 input dims_y\n",
    "k2 = 128        # num lvl2 filters\n",
    "stride_y = 2    # lvl2 stride"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Third layer:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# variables for third layer\n",
    "d3_x = 25 # lvl3 input dims_x\n",
    "d3_y = 1 # lvl3 input dims_y (fully connected)\n",
    "k3 = 4096 # num lvl3 filters"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Calculate the sizes of nodes at each layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "First layer regions: (25,512)\n",
      "Second layer regions: (25,193)\n",
      "Third layer regions: (1,193)\n"
     ]
    }
   ],
   "source": [
    "num_regions  = int(1 + (window-dd)/stride)\n",
    "print ('First layer regions: ({},{})'.format(num_regions,kk))\n",
    "num_regions2_x  = int(1 + (num_regions-d2_x)/1)\n",
    "num_regions2_y = int(1 + (kk-d2_y)/stride_y)\n",
    "print ('Second layer regions: ({},{})'.format(num_regions2_x,num_regions2_y))\n",
    "num_regions3_x = int(1 + (num_regions2_x - d3_x)/1)\n",
    "num_regions3_y = int(1 + (num_regions2_y - d3_y)/1)\n",
    "print ('Third layer regions: ({},{})'.format(num_regions3_x,num_regions3_y))\n",
    "\n",
    "# create filters (shape: (1, 4096, 1, 512)) = number of regions x size of region\n",
    "wsin,wcos = create_filters(dd,kk)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5.2 Define tf.Graph()\n",
    "Set it all together and define the graph.\n",
    "\n",
    "For evaluation not the trained weights, but its moving averages with decay 0.9998 are utilized. The shadow variables are maintained by tf.train.ExponentialMovingAverage. In the block for evaluating the direct model their values are retrieved."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---- Weights ----\n",
      "w <tf.Variable 'parameters/w:0' shape=(1, 128, 1, 128) dtype=float32_ref>\n",
      "w2 <tf.Variable 'parameters/w2:0' shape=(25, 1, 128, 4096) dtype=float32_ref>\n",
      "beta <tf.Variable 'parameters/beta:0' shape=(790529, 128) dtype=float32_ref>\n",
      "---- Layers ----\n",
      "zx Tensor(\"queued_model/add:0\", shape=(?, 1, 25, 512), dtype=float32)\n",
      "z2 Tensor(\"queued_model/Relu:0\", shape=(?, 128, 25, 193), dtype=float32)\n",
      "z3 Tensor(\"queued_model/Relu_1:0\", shape=(?, 4096, 1, 193), dtype=float32)\n",
      "y Tensor(\"queued_model/MatMul:0\", shape=(150, 128), dtype=float32)\n",
      "loss Tensor(\"queued_model/logistic_loss:0\", shape=(150, 128), dtype=float32)\n"
     ]
    }
   ],
   "source": [
    "tf.set_random_seed(999)\n",
    "\n",
    "#end_learning_rate = .00001/3/3/3\n",
    "starter_learning_rate = .005\n",
    "mom = .95\n",
    "\n",
    "beta_reg = 0.001\n",
    "\n",
    "graph = tf.Graph()\n",
    "\n",
    "with graph.as_default():\n",
    "     \n",
    "    # if the input is a batch\n",
    "    with tf.variable_scope('data_queue'):\n",
    "        xb = tf.placeholder(tf.float32, shape=[None, 1, window+(out-1)*stride, 1])\n",
    "        yb = tf.placeholder(tf.float32, shape=[None, out, mm])\n",
    "\n",
    "        xq, yq = xb, yb\n",
    "        \n",
    "    # if the input is just one segment\n",
    "    with tf.variable_scope('direct_data'):\n",
    "        xd = tf.placeholder(tf.float32, shape=[None,1,window,1])\n",
    "        yd = tf.placeholder(tf.float32, shape=[None, mm])\n",
    "    \n",
    "    print ('---- Weights ----')\n",
    "    with tf.variable_scope('parameters'): #glorot  \n",
    "        w = tf.get_variable('w', [d2_x,d2_y,1,k2]) # 1 x 128 x 1 x 128 (16.384 values)\n",
    "        print ('w',w)\n",
    "        w2 = tf.get_variable('w2', [d3_x,d3_y,k2,k3]) # 25 x 1 x 128 x 4096 (13.107.200 values)\n",
    "        print ('w2',w2)\n",
    "        beta = tf.get_variable('beta', [int(num_regions3_x*num_regions3_y*k3)+1,mm]) \n",
    "        # (1*193*4096+1) x 128 (101.187.584 + 128 values) BIAS!!\n",
    "        print ('beta',beta)\n",
    "        model_stats.register_weights([w,w2,beta],['w','w2','beta'])\n",
    "\n",
    "    print ('---- Layers ----')\n",
    "    with tf.variable_scope('queued_model'):\n",
    "        zx = tf.square(tf.nn.conv2d(xq,wsin,strides=[1,1,stride,1],padding='VALID')) \\\n",
    "           + tf.square(tf.nn.conv2d(xq,wcos,strides=[1,1,stride,1],padding='VALID'))\n",
    "        print ('zx',zx)\n",
    "        z2 = tf.nn.relu(tf.nn.conv2d(tf.log(zx+10e-15),w,strides=[1,1,1,stride_y],padding='VALID',data_format='NCHW'))\n",
    "        print ('z2',z2)\n",
    "        z3 = tf.nn.relu(tf.nn.conv2d(z2,w2,strides=[1,1,1,1],padding='VALID',data_format='NCHW'))\n",
    "        print ('z3',z3)\n",
    "        y = tf.matmul(\n",
    "            tf.concat([tf.reshape(z3,[batch_size,int(num_regions3_x*num_regions3_y*k3)]),\n",
    "                      tf.ones([batch_size,1], tf.float32)],1),beta)\n",
    "        print ('y',y)\n",
    "        #loss = tf.reduce_mean(tf.nn.l2_loss(y-tf.reshape(yq,[batch_size,mm])))\n",
    "        loss = tf.nn.sigmoid_cross_entropy_with_logits(labels=tf.reshape(yq,[batch_size,mm]), logits=y) \n",
    "        print ('loss',loss)\n",
    "            \n",
    "    # learning rate decay\n",
    "    global_step = tf.Variable(0, trainable=False)\n",
    "    lr = tf.train.exponential_decay(starter_learning_rate, global_step, 10000, 0.95, staircase=True)\n",
    "\n",
    "    reg = tf.nn.l2_loss(w) + tf.nn.l2_loss(w2) + tf.nn.l2_loss(beta)\n",
    "    \n",
    "    opt_op = tf.train.MomentumOptimizer(lr,mom).minimize(tf.reduce_mean(loss + beta_reg*reg), global_step=global_step)\n",
    "    \n",
    "    #opt_op = tf.train.MomentumOptimizer(lr,mom).minimize(tf.reduce_mean(loss), global_step=global_step)\n",
    "\n",
    "    \n",
    "    # Create an ExponentialMovingAverage object\n",
    "    ema = tf.train.ExponentialMovingAverage(decay=0.9998)\n",
    "    \n",
    "    with tf.control_dependencies([opt_op]):\n",
    "        training_op = ema.apply([w, w2, beta])\n",
    "    \n",
    "    with tf.variable_scope('direct_model'):\n",
    "        # use ema.average(var) to return the average of variable var\n",
    "        wavg = ema.average(w)\n",
    "        w2avg = ema.average(w2)\n",
    "        betaavg = ema.average(beta)\n",
    "\n",
    "        direct_zx = tf.square(tf.nn.conv2d(xd,wsin,strides=[1,1,stride,1],padding='VALID')) \\\n",
    "                + tf.square(tf.nn.conv2d(xd,wcos,strides=[1,1,stride,1],padding='VALID'))\n",
    "        direct_z2 = tf.nn.relu(tf.nn.conv2d(tf.log(direct_zx+10e-15),wavg,strides=[1,1,1,stride_y],padding='VALID',data_format='NCHW'))\n",
    "        direct_z3 = tf.nn.relu(tf.nn.conv2d(direct_z2,w2avg,strides=[1,1,1,1],padding='VALID',data_format='NCHW'))\n",
    "        direct_y = tf.nn.sigmoid(tf.matmul(\n",
    "            tf.concat([tf.reshape(direct_z3,[tf.shape(xd)[0],int(num_regions3_x*num_regions3_y*k3)]),\n",
    "                     tf.ones([tf.shape(xd)[0],1], tf.float32)],1),betaavg))\n",
    "        direct_loss = tf.reduce_mean(tf.nn.l2_loss(direct_y-yd))\n",
    "        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5.3 Evaluation of the graph\n",
    "\n",
    "For the aplication of the neuronal network a moving average of the weights is used. Therefore, a different function for evaluation is used.\n",
    "\n",
    "Inside the function predict_direct_model the mean squared error is calculated directly."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# function for prediction without training that splits up the whole set\n",
    "# in small subsets to prevent errors of type \"ResourceExhaustedError\"\n",
    "# input: \n",
    "# - sess: the actual session\n",
    "# - X: sample, size: (n,1,window,1)\n",
    "# - Y: corresponding labels, size: (n,mm)\n",
    "# output:\n",
    "#  Yhat: prediction, size: (n,mm)\n",
    "#  mse: Mean Squared Error when comparing Y and Yhat\n",
    "def predict_direct_model(sess, X, Y):\n",
    "    Yhat = np.empty((len(X),mm))\n",
    "    subdiv = 100\n",
    "    subset = X.shape[0]/subdiv\n",
    "    mse = 0\n",
    "    for j in range(subdiv):\n",
    "        Yhat[int(subset*j):int(subset*(j+1))], se = sess.run([direct_y,direct_loss], \n",
    "                                                         feed_dict={xd: X[int(subset*j):int(subset*(j+1))],\n",
    "                                                                    yd: Y[int(subset*j):int(subset*(j+1))]})\n",
    "        mse += se\n",
    "\n",
    "    return Yhat, mse/float(subdiv)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6. Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "if debugging == True:\n",
    "    num_steps = 5\n",
    "    control_step = 1\n",
    "else:\n",
    "    num_steps = 300000\n",
    "    control_step = 1000\n",
    "\n",
    "gpu_memory_growth = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# arrays for input data (batch)\n",
    "xmb = np.empty([batch_size,1,window,1],dtype=np.float32)\n",
    "ymb = np.empty([batch_size,1,mm],dtype=np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train saver\n",
    "saver = tf.train.Saver({'wavg': wavg, 'w2avg': w2avg, 'betaavg': betaavg})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized\n",
      "iter\t test_squ_loss\t test_avg_prec\t train_squ_loss\t train_avg_prec\t regularization\t time (in s)\n",
      "1 \t 1759.24099365 \t 0.01829637 \t 2397.60881348 \t 0.02447197 \t 253.11 \t 43.96683335\n",
      "1001 \t 628.95217133 \t 0.26540181 \t 911.46930908 \t 0.32198955 \t 210.339 \t 653.33821988\n",
      "2001 \t 319.94045532 \t 0.50366021 \t 480.32232513 \t 0.54156016 \t 173.078 \t 648.9419744\n",
      "3001 \t 196.56282867 \t 0.58286275 \t 301.47463531 \t 0.60957113 \t 142.466 \t 652.41518116\n",
      "4001 \t 135.86790009 \t 0.61845684 \t 212.23669174 \t 0.63765514 \t 117.401 \t 650.83550286\n",
      "5001 \t 104.19691124 \t 0.63838904 \t 165.58737244 \t 0.65242397 \t 96.8605 \t 639.52318954\n",
      "6001 \t 87.31431416 \t 0.65083502 \t 140.80395256 \t 0.66138698 \t 80.0343 \t 649.62681103\n",
      "7001 \t 77.96727203 \t 0.6592902 \t 127.22979622 \t 0.66740739 \t 66.2599 \t 644.32499909\n",
      "8001 \t 72.59596333 \t 0.66519465 \t 119.53448059 \t 0.67169543 \t 54.987 \t 638.53809214\n",
      "9001 \t 69.37192528 \t 0.66956518 \t 115.00602081 \t 0.67485923 \t 45.7402 \t 649.69667578\n",
      "10001 \t 67.34565651 \t 0.67306799 \t 112.25126274 \t 0.67726662 \t 38.1704 \t 649.38743949\n",
      "11001 \t 66.04183565 \t 0.67573951 \t 110.51466614 \t 0.67923393 \t 32.2485 \t 638.13488698\n",
      "12001 \t 65.15967848 \t 0.67780905 \t 109.37256142 \t 0.68084849 \t 27.3582 \t 646.56769562\n",
      "13001 \t 64.55253672 \t 0.67958394 \t 108.59839668 \t 0.68224032 \t 23.3013 \t 652.13441062\n",
      "14001 \t 64.11228178 \t 0.68115591 \t 108.03195297 \t 0.68353354 \t 19.9399 \t 652.95850468\n",
      "15001 \t 63.77505173 \t 0.68255692 \t 107.61258263 \t 0.68467278 \t 17.1716 \t 653.51707768\n",
      "16001 \t 63.50263311 \t 0.68381827 \t 107.26643188 \t 0.68574313 \t 14.8801 \t 647.09364462\n",
      "17001 \t 63.27248384 \t 0.68507735 \t 106.98325302 \t 0.68666684 \t 12.973 \t 642.85064483\n",
      "18001 \t 63.07274292 \t 0.68629956 \t 106.73481255 \t 0.68756103 \t 11.3954 \t 643.40370011\n",
      "19001 \t 62.90628925 \t 0.68728392 \t 106.50768608 \t 0.68841429 \t 10.0928 \t 645.28368878\n",
      "20001 \t 62.76960602 \t 0.68813731 \t 106.30767548 \t 0.68917753 \t 9.01549 \t 660.38020849\n",
      "21001 \t 62.65209431 \t 0.68900182 \t 106.13267822 \t 0.68985929 \t 8.15055 \t 651.1910398\n",
      "22001 \t 62.54246115 \t 0.68969508 \t 105.96733391 \t 0.69053762 \t 7.437 \t 647.82347822\n",
      "23001 \t 62.44396025 \t 0.69036655 \t 105.80050507 \t 0.69122447 \t 6.84352 \t 638.66978335\n",
      "24001 \t 62.35465868 \t 0.69095069 \t 105.64529808 \t 0.69184066 \t 6.34773 \t 644.69647336\n",
      "25001 \t 62.25177303 \t 0.69169571 \t 105.48519829 \t 0.69249581 \t 5.93203 \t 657.72627544\n",
      "26001 \t 62.17775421 \t 0.69224727 \t 105.33949249 \t 0.69311921 \t 5.57131 \t 651.11404181\n",
      "27001 \t 62.08265711 \t 0.69295826 \t 105.19523361 \t 0.69371146 \t 5.29023 \t 646.93697691\n",
      "28001 \t 62.00352016 \t 0.69356603 \t 105.05989632 \t 0.69425572 \t 5.04372 \t 637.83395219\n",
      "29001 \t 61.94368847 \t 0.69411626 \t 104.93950691 \t 0.69475071 \t 4.84336 \t 641.05423856\n",
      "30001 \t 61.86828072 \t 0.69461648 \t 104.80828415 \t 0.69523828 \t 4.67126 \t 650.13067985\n",
      "31001 \t 61.80905365 \t 0.69506295 \t 104.69341721 \t 0.69567939 \t 4.53769 \t 654.90489292\n",
      "32001 \t 61.73471487 \t 0.69563931 \t 104.58143265 \t 0.69614067 \t 4.42312 \t 643.02443576\n",
      "33001 \t 61.69276415 \t 0.69599525 \t 104.48339127 \t 0.69654477 \t 4.33345 \t 643.21022677\n",
      "34001 \t 61.63255955 \t 0.69642662 \t 104.37622017 \t 0.69701364 \t 4.24579 \t 642.34290075\n",
      "35001 \t 61.57437767 \t 0.69685032 \t 104.28792343 \t 0.69740888 \t 4.17536 \t 653.85367751\n",
      "36001 \t 61.51651426 \t 0.69736917 \t 104.19111557 \t 0.69786941 \t 4.12193 \t 656.48095155\n",
      "37001 \t 61.44800013 \t 0.69791268 \t 104.10643799 \t 0.69829469 \t 4.07697 \t 649.10290265\n",
      "38001 \t 61.39920553 \t 0.6983049 \t 104.02224991 \t 0.69864677 \t 4.03766 \t 641.61620712\n",
      "39001 \t 61.34150051 \t 0.69878897 \t 103.94423988 \t 0.69899057 \t 4.01518 \t 648.97791457\n",
      "40001 \t 61.2799745 \t 0.69926029 \t 103.86371849 \t 0.69932813 \t 3.97519 \t 652.21290421\n",
      "41001 \t 61.23880766 \t 0.69964988 \t 103.78758041 \t 0.69967412 \t 3.94762 \t 660.28316355\n",
      "42001 \t 61.17839911 \t 0.7002105 \t 103.7126738 \t 0.69999924 \t 3.9314 \t 655.44087005\n",
      "43001 \t 61.13905666 \t 0.70057943 \t 103.65844742 \t 0.70025225 \t 3.92165 \t 650.41834092\n",
      "44001 \t 61.08852488 \t 0.70098835 \t 103.60853203 \t 0.70048668 \t 3.90728 \t 646.04059768\n",
      "45001 \t 61.04384695 \t 0.70137045 \t 103.54300644 \t 0.70073474 \t 3.89743 \t 646.62266088\n",
      "46001 \t 61.04039928 \t 0.70149168 \t 103.48868576 \t 0.70098941 \t 3.88342 \t 649.34575462\n",
      "47001 \t 61.02915514 \t 0.70154684 \t 103.43389359 \t 0.7012518 \t 3.88236 \t 659.28856301\n",
      "48001 \t 61.00956829 \t 0.70172734 \t 103.38124748 \t 0.70154867 \t 3.87035 \t 648.88261819\n",
      "49001 \t 60.97041105 \t 0.7020856 \t 103.31847618 \t 0.70187531 \t 3.86114 \t 640.40817952\n",
      "50001 \t 60.94354002 \t 0.70237997 \t 103.27067467 \t 0.70213879 \t 3.85914 \t 649.68032551\n",
      "51001 \t 60.93020164 \t 0.70249289 \t 103.23025017 \t 0.70235951 \t 3.85684 \t 658.60782027\n",
      "52001 \t 60.89252123 \t 0.70283333 \t 103.17869621 \t 0.70261624 \t 3.85578 \t 654.8052125\n",
      "53001 \t 60.85418795 \t 0.70317581 \t 103.1385791 \t 0.70281048 \t 3.8552 \t 639.75752211\n",
      "54001 \t 60.84175089 \t 0.7032351 \t 103.09920372 \t 0.70298139 \t 3.84345 \t 656.10219193\n",
      "55001 \t 60.80389339 \t 0.70358606 \t 103.05042061 \t 0.70320955 \t 3.84641 \t 663.60778785\n",
      "56001 \t 60.80239882 \t 0.70356705 \t 103.02104126 \t 0.70336889 \t 3.84288 \t 669.33913159\n",
      "57001 \t 60.78825979 \t 0.703694 \t 102.99326614 \t 0.70349351 \t 3.83718 \t 663.26526785\n",
      "58001 \t 60.76051222 \t 0.70394205 \t 102.95856346 \t 0.70365823 \t 3.83627 \t 656.22961593\n",
      "59001 \t 60.74668478 \t 0.70398278 \t 102.92337334 \t 0.70385626 \t 3.84122 \t 657.27115321\n",
      "60001 \t 60.72665321 \t 0.70420203 \t 102.89353851 \t 0.70399935 \t 3.83501 \t 652.63528275\n",
      "61001 \t 60.70555635 \t 0.7044 \t 102.86379128 \t 0.70414143 \t 3.83274 \t 653.89801741\n",
      "62001 \t 60.69342844 \t 0.70456029 \t 102.83681213 \t 0.70424008 \t 3.82772 \t 666.21633315\n",
      "63001 \t 60.68963575 \t 0.70464744 \t 102.80493156 \t 0.70440256 \t 3.83213 \t 658.04572082\n",
      "64001 \t 60.67787539 \t 0.70480668 \t 102.78584991 \t 0.70451813 \t 3.8275 \t 656.79660225\n",
      "65001 \t 60.67642929 \t 0.70479729 \t 102.77024117 \t 0.70455716 \t 3.82371 \t 643.33551288\n",
      "66001 \t 60.64224615 \t 0.70509824 \t 102.74965775 \t 0.70464005 \t 3.82873 \t 652.20613337\n",
      "67001 \t 60.62170553 \t 0.70521629 \t 102.72124184 \t 0.7047719 \t 3.82332 \t 657.6143074\n",
      "68001 \t 60.5974358 \t 0.70531893 \t 102.71287025 \t 0.7048226 \t 3.82106 \t 658.24829459\n",
      "69001 \t 60.58697733 \t 0.70547286 \t 102.70878983 \t 0.70488207 \t 3.81717 \t 643.62470746\n",
      "70001 \t 60.56607779 \t 0.70564732 \t 102.70344757 \t 0.70488377 \t 3.82831 \t 645.22885966\n",
      "71001 \t 60.58300247 \t 0.70554771 \t 102.68200806 \t 0.70496527 \t 3.82222 \t 654.80327272\n",
      "72001 \t 60.57199415 \t 0.70566753 \t 102.66940887 \t 0.70497961 \t 3.82732 \t 656.82043123\n",
      "73001 \t 60.56354462 \t 0.70575665 \t 102.66197876 \t 0.70498971 \t 3.8217 \t 647.74232888\n",
      "74001 \t 60.54820583 \t 0.70588737 \t 102.6397892 \t 0.7050897 \t 3.81962 \t 652.52283978\n",
      "75001 \t 60.54361273 \t 0.70590148 \t 102.61835495 \t 0.70520043 \t 3.82109 \t 649.11048388\n",
      "76001 \t 60.53255531 \t 0.70591145 \t 102.60132729 \t 0.7052757 \t 3.82052 \t 639.82354355\n",
      "77001 \t 60.53355011 \t 0.7058903 \t 102.58483177 \t 0.70538442 \t 3.82228 \t 646.19204307\n",
      "78001 \t 60.5144022 \t 0.70600411 \t 102.56371094 \t 0.70549307 \t 3.81795 \t 647.8165102\n",
      "79001 \t 60.50625822 \t 0.70603441 \t 102.54073799 \t 0.70561151 \t 3.82324 \t 626.82069612\n",
      "80001 \t 60.4786095 \t 0.70632011 \t 102.53333824 \t 0.70565528 \t 3.81918 \t 632.46366119\n",
      "81001 \t 60.47033714 \t 0.70632329 \t 102.50952927 \t 0.70577056 \t 3.82136 \t 633.45651603\n",
      "82001 \t 60.46127243 \t 0.70644493 \t 102.49792839 \t 0.70578695 \t 3.82467 \t 639.90193009\n",
      "83001 \t 60.44021374 \t 0.7066843 \t 102.47048706 \t 0.7059003 \t 3.81849 \t 642.82229114\n",
      "84001 \t 60.42299713 \t 0.70679406 \t 102.45396484 \t 0.70597412 \t 3.82482 \t 639.74039149\n",
      "85001 \t 60.40912149 \t 0.70693935 \t 102.44493401 \t 0.70601416 \t 3.8221 \t 638.99588346\n",
      "86001 \t 60.401381 \t 0.70712424 \t 102.43756775 \t 0.70608994 \t 3.81958 \t 632.3615098\n",
      "87001 \t 60.38288473 \t 0.70726428 \t 102.4195018 \t 0.70620004 \t 3.82326 \t 636.49637318\n",
      "88001 \t 60.34482851 \t 0.70760378 \t 102.40334496 \t 0.70627358 \t 3.81677 \t 641.98992634\n",
      "89001 \t 60.34169979 \t 0.70766626 \t 102.3811615 \t 0.70639525 \t 3.81954 \t 641.79277205\n",
      "90001 \t 60.34996756 \t 0.70757351 \t 102.369459 \t 0.70645238 \t 3.81507 \t 636.19413972\n",
      "91001 \t 60.32827812 \t 0.7076181 \t 102.34884163 \t 0.70656906 \t 3.81629 \t 623.8963294\n",
      "92001 \t 60.32535034 \t 0.70764793 \t 102.33941376 \t 0.70661858 \t 3.80831 \t 631.77961946\n",
      "93001 \t 60.32524467 \t 0.70778333 \t 102.32698318 \t 0.70673311 \t 3.81591 \t 641.53408456\n",
      "94001 \t 60.31224535 \t 0.70792095 \t 102.31545151 \t 0.70678453 \t 3.81522 \t 644.22772884\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "95001 \t 60.28318367 \t 0.7081997 \t 102.29875648 \t 0.706871 \t 3.81183 \t 639.11227179\n",
      "96001 \t 60.28574011 \t 0.70814778 \t 102.29892097 \t 0.70684013 \t 3.81528 \t 630.94154263\n",
      "97001 \t 60.27779097 \t 0.70814532 \t 102.28843689 \t 0.70687542 \t 3.81301 \t 640.779531\n",
      "98001 \t 60.27413834 \t 0.70811576 \t 102.27941971 \t 0.70688937 \t 3.81085 \t 641.92598915\n",
      "99001 \t 60.27495789 \t 0.70804854 \t 102.26910286 \t 0.70696008 \t 3.80941 \t 643.93874979\n",
      "100001 \t 60.27838058 \t 0.70801774 \t 102.25692757 \t 0.7070047 \t 3.81013 \t 642.76475072\n",
      "101001 \t 60.28650156 \t 0.7079158 \t 102.24488213 \t 0.70706258 \t 3.81087 \t 630.92592382\n",
      "102001 \t 60.29403599 \t 0.70789353 \t 102.2295443 \t 0.70712779 \t 3.80901 \t 628.41129827\n",
      "103001 \t 60.28199287 \t 0.70805852 \t 102.21171829 \t 0.70722373 \t 3.80199 \t 639.29724193\n",
      "104001 \t 60.27222692 \t 0.70820227 \t 102.21350418 \t 0.7072304 \t 3.80904 \t 646.90667343\n",
      "105001 \t 60.24377537 \t 0.7084034 \t 102.20403595 \t 0.70729102 \t 3.80307 \t 644.38879704\n",
      "106001 \t 60.22922958 \t 0.70854051 \t 102.19203873 \t 0.70731416 \t 3.80687 \t 641.88722301\n",
      "107001 \t 60.22020948 \t 0.708633 \t 102.18383148 \t 0.70736326 \t 3.80498 \t 629.02559638\n",
      "108001 \t 60.21642988 \t 0.7086389 \t 102.17359604 \t 0.70739391 \t 3.80624 \t 633.96255946\n",
      "109001 \t 60.21045914 \t 0.70867883 \t 102.16475243 \t 0.70748807 \t 3.80387 \t 642.709831\n",
      "110001 \t 60.20760677 \t 0.70868644 \t 102.15380096 \t 0.70757767 \t 3.8069 \t 651.88113236\n",
      "111001 \t 60.22050449 \t 0.70857948 \t 102.14110054 \t 0.70764565 \t 3.80583 \t 640.91795969\n",
      "112001 \t 60.21989445 \t 0.7086464 \t 102.13764603 \t 0.70765014 \t 3.80274 \t 623.5348773\n",
      "113001 \t 60.22679041 \t 0.70868619 \t 102.13870567 \t 0.70768048 \t 3.79573 \t 631.75209308\n",
      "114001 \t 60.23236338 \t 0.70859858 \t 102.13466461 \t 0.70769952 \t 3.80345 \t 636.46605325\n",
      "115001 \t 60.23433573 \t 0.70856347 \t 102.13488625 \t 0.70771783 \t 3.79606 \t 640.47208238\n",
      "116001 \t 60.228724 \t 0.70862595 \t 102.13095535 \t 0.70772214 \t 3.80131 \t 638.68864202\n",
      "117001 \t 60.23218128 \t 0.70853048 \t 102.11594826 \t 0.70776345 \t 3.79943 \t 636.50683379\n",
      "118001 \t 60.24452293 \t 0.70842819 \t 102.11130196 \t 0.70777909 \t 3.79574 \t 634.10723567\n",
      "119001 \t 60.25214602 \t 0.70845279 \t 102.10636124 \t 0.70780753 \t 3.79328 \t 632.74239683\n",
      "120001 \t 60.24718126 \t 0.70846905 \t 102.10876526 \t 0.70781777 \t 3.79548 \t 647.91885877\n",
      "121001 \t 60.25379288 \t 0.70847482 \t 102.12017365 \t 0.70779299 \t 3.79555 \t 632.81586027\n",
      "122001 \t 60.24037022 \t 0.70855528 \t 102.10863403 \t 0.70786844 \t 3.791 \t 639.85885859\n",
      "123001 \t 60.21728691 \t 0.70871641 \t 102.09987022 \t 0.7078984 \t 3.79261 \t 630.71817207\n",
      "124001 \t 60.20324789 \t 0.70878802 \t 102.09473145 \t 0.70792536 \t 3.79088 \t 632.98292327\n",
      "125001 \t 60.20930761 \t 0.70874387 \t 102.08086784 \t 0.7080092 \t 3.79141 \t 633.969877\n",
      "126001 \t 60.20739542 \t 0.70864515 \t 102.08229256 \t 0.70797607 \t 3.78809 \t 640.93834257\n",
      "127001 \t 60.20651079 \t 0.70853987 \t 102.07749275 \t 0.70799562 \t 3.79147 \t 645.19801855\n",
      "128001 \t 60.18487909 \t 0.7086657 \t 102.07399055 \t 0.7079891 \t 3.79071 \t 632.60853815\n",
      "129001 \t 60.20186365 \t 0.70848349 \t 102.07396126 \t 0.70798851 \t 3.7901 \t 625.08062649\n",
      "130001 \t 60.20922409 \t 0.70841744 \t 102.06391769 \t 0.70800617 \t 3.78968 \t 636.46039605\n",
      "131001 \t 60.18096481 \t 0.7086615 \t 102.04988625 \t 0.70808059 \t 3.79118 \t 630.04294777\n",
      "132001 \t 60.1766987 \t 0.7087084 \t 102.03232269 \t 0.70812291 \t 3.7856 \t 645.15669537\n",
      "133001 \t 60.17226931 \t 0.70865511 \t 102.01570168 \t 0.70819079 \t 3.78548 \t 634.30467391\n",
      "134001 \t 60.15126144 \t 0.70887628 \t 102.00959236 \t 0.7082378 \t 3.78568 \t 624.80559063\n",
      "135001 \t 60.13977259 \t 0.7090126 \t 101.99976189 \t 0.70827781 \t 3.78495 \t 632.88778234\n",
      "136001 \t 60.14741695 \t 0.70897353 \t 102.00471832 \t 0.70827931 \t 3.79088 \t 660.23032188\n",
      "137001 \t 60.15512814 \t 0.70897126 \t 102.00575035 \t 0.70826883 \t 3.79209 \t 632.48619556\n",
      "138001 \t 60.17375692 \t 0.70888058 \t 102.00918144 \t 0.7082413 \t 3.78987 \t 631.83845711\n",
      "139001 \t 60.16340458 \t 0.70897068 \t 102.01149292 \t 0.70823197 \t 3.78211 \t 633.76706386\n",
      "140001 \t 60.16289957 \t 0.70895447 \t 102.00738129 \t 0.70824438 \t 3.78569 \t 632.36156702\n",
      "141001 \t 60.14006315 \t 0.70915585 \t 102.01988846 \t 0.70817737 \t 3.78389 \t 637.5417459\n",
      "142001 \t 60.13889736 \t 0.70909565 \t 102.01833252 \t 0.70815675 \t 3.78814 \t 643.93034792\n",
      "143001 \t 60.13988445 \t 0.70905403 \t 102.01817299 \t 0.70814529 \t 3.78628 \t 630.64631629\n",
      "144001 \t 60.14196735 \t 0.70907578 \t 102.01294449 \t 0.70817696 \t 3.7889 \t 627.23868656\n",
      "145001 \t 60.13506086 \t 0.70922469 \t 102.00912308 \t 0.70823008 \t 3.78303 \t 635.3622725\n",
      "146001 \t 60.1189114 \t 0.709345 \t 102.00635757 \t 0.70822426 \t 3.78195 \t 645.11698151\n",
      "147001 \t 60.12581375 \t 0.70934385 \t 102.00620232 \t 0.70824026 \t 3.78421 \t 653.1526587\n",
      "148001 \t 60.11398874 \t 0.70946598 \t 102.0026873 \t 0.7082693 \t 3.78788 \t 640.90892768\n",
      "149001 \t 60.11449757 \t 0.70945909 \t 102.00316154 \t 0.70830449 \t 3.78163 \t 640.61810994\n",
      "150001 \t 60.12388939 \t 0.70935621 \t 102.01806488 \t 0.70822889 \t 3.77721 \t 631.87621284\n",
      "151001 \t 60.13992069 \t 0.70921726 \t 102.01480705 \t 0.70821347 \t 3.78228 \t 634.76626492\n",
      "152001 \t 60.11426334 \t 0.7093811 \t 102.00794037 \t 0.70820314 \t 3.7824 \t 642.03569055\n",
      "153001 \t 60.11215952 \t 0.70940305 \t 102.00984497 \t 0.70819842 \t 3.77634 \t 642.62923908\n",
      "154001 \t 60.11508146 \t 0.70936912 \t 102.01495247 \t 0.70819734 \t 3.78172 \t 637.71657944\n",
      "155001 \t 60.13161938 \t 0.70926827 \t 102.00884743 \t 0.70822524 \t 3.77931 \t 621.4028554\n",
      "156001 \t 60.12320427 \t 0.7092849 \t 101.99542191 \t 0.70828606 \t 3.77765 \t 635.48967528\n",
      "157001 \t 60.11984713 \t 0.70926722 \t 101.9873674 \t 0.70832875 \t 3.78192 \t 637.91285133\n",
      "158001 \t 60.09147259 \t 0.70955918 \t 101.98163857 \t 0.70833896 \t 3.77859 \t 643.82638407\n",
      "159001 \t 60.08936188 \t 0.70957859 \t 101.98103485 \t 0.70833886 \t 3.77618 \t 636.43377352\n",
      "160001 \t 60.09810007 \t 0.70946386 \t 101.97878967 \t 0.70833301 \t 3.77924 \t 627.34296703\n",
      "161001 \t 60.09889149 \t 0.70946463 \t 101.97051521 \t 0.70833565 \t 3.77644 \t 637.00164032\n",
      "162001 \t 60.08956142 \t 0.70946069 \t 101.96301277 \t 0.70831763 \t 3.78152 \t 636.314785\n",
      "163001 \t 60.08259737 \t 0.70950349 \t 101.9529509 \t 0.70834312 \t 3.77513 \t 643.92898989\n",
      "164001 \t 60.05978815 \t 0.70961887 \t 101.9487191 \t 0.70837058 \t 3.77481 \t 634.2301836\n",
      "165001 \t 60.03294302 \t 0.70977365 \t 101.93872742 \t 0.70844578 \t 3.77772 \t 635.32049155\n",
      "166001 \t 60.04159796 \t 0.7097354 \t 101.93229523 \t 0.70843476 \t 3.77996 \t 634.1086607\n",
      "167001 \t 60.03361172 \t 0.70983909 \t 101.92390076 \t 0.70844229 \t 3.7734 \t 635.9487505\n",
      "168001 \t 60.02691544 \t 0.70992488 \t 101.9228582 \t 0.70845397 \t 3.77483 \t 644.16268373\n",
      "169001 \t 60.0295837 \t 0.70985652 \t 101.92308372 \t 0.70845231 \t 3.77333 \t 638.77065682\n",
      "170001 \t 60.01897364 \t 0.70993571 \t 101.92296097 \t 0.70844387 \t 3.77332 \t 640.46667576\n",
      "171001 \t 60.02757915 \t 0.70989201 \t 101.92468903 \t 0.70842735 \t 3.77451 \t 629.95840144\n",
      "172001 \t 60.03382702 \t 0.70986755 \t 101.93182602 \t 0.70839084 \t 3.77308 \t 633.15878296\n",
      "173001 \t 60.06017456 \t 0.70972496 \t 101.92368851 \t 0.7084349 \t 3.77606 \t 638.99682355\n",
      "174001 \t 60.07984371 \t 0.70961274 \t 101.92190849 \t 0.70844935 \t 3.77421 \t 632.27717352\n",
      "175001 \t 60.06654613 \t 0.70974185 \t 101.9185778 \t 0.70843219 \t 3.77666 \t 636.32064939\n",
      "176001 \t 60.06280035 \t 0.70981714 \t 101.9138765 \t 0.70848478 \t 3.77125 \t 626.62904692\n",
      "177001 \t 60.05036961 \t 0.70993847 \t 101.9106076 \t 0.70852632 \t 3.77715 \t 624.17898726\n",
      "178001 \t 60.05328205 \t 0.70988928 \t 101.90384209 \t 0.70859131 \t 3.77668 \t 639.94754219\n",
      "179001 \t 60.08415079 \t 0.70964303 \t 101.89237381 \t 0.70864587 \t 3.77232 \t 636.06174231\n",
      "180001 \t 60.09955963 \t 0.7094668 \t 101.90148659 \t 0.70860711 \t 3.77253 \t 634.74672055\n",
      "181001 \t 60.11365217 \t 0.70941492 \t 101.89738029 \t 0.70864469 \t 3.77273 \t 635.04597044\n",
      "182001 \t 60.12662468 \t 0.70930246 \t 101.89047958 \t 0.70869318 \t 3.77214 \t 635.44459701\n",
      "183001 \t 60.12467087 \t 0.70925671 \t 101.87797142 \t 0.70872651 \t 3.77441 \t 644.47768569\n",
      "184001 \t 60.12061134 \t 0.70940839 \t 101.86690254 \t 0.70878693 \t 3.77479 \t 639.41109371\n",
      "185001 \t 60.13581919 \t 0.709248 \t 101.8588765 \t 0.7088127 \t 3.77448 \t 645.0492425\n",
      "186001 \t 60.11699249 \t 0.70936003 \t 101.85257652 \t 0.70884055 \t 3.77284 \t 638.3408525\n",
      "187001 \t 60.11106388 \t 0.70939419 \t 101.84628517 \t 0.70888362 \t 3.77257 \t 630.3600266\n",
      "188001 \t 60.10799618 \t 0.70943937 \t 101.84028534 \t 0.70892437 \t 3.77035 \t 631.11814022\n",
      "189001 \t 60.11005424 \t 0.70944658 \t 101.83876625 \t 0.70892027 \t 3.76885 \t 645.56207681\n",
      "190001 \t 60.1192362 \t 0.70935532 \t 101.83927292 \t 0.70890586 \t 3.77365 \t 639.03600693\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "191001 \t 60.09619497 \t 0.70948906 \t 101.83551292 \t 0.70891945 \t 3.76937 \t 644.34133577\n",
      "192001 \t 60.08106863 \t 0.70956533 \t 101.8316095 \t 0.70893841 \t 3.76684 \t 635.38992405\n",
      "193001 \t 60.07485285 \t 0.70966314 \t 101.83043022 \t 0.70896628 \t 3.76267 \t 627.89405322\n",
      "194001 \t 60.08823677 \t 0.70957709 \t 101.82961456 \t 0.70894828 \t 3.76653 \t 632.95961094\n",
      "195001 \t 60.09002707 \t 0.70968197 \t 101.82980911 \t 0.70895146 \t 3.76616 \t 644.80582952\n",
      "196001 \t 60.09846085 \t 0.70964722 \t 101.83493629 \t 0.70893141 \t 3.76644 \t 643.48635578\n",
      "197001 \t 60.12253881 \t 0.70946212 \t 101.84020912 \t 0.70888216 \t 3.76798 \t 631.19277954\n",
      "198001 \t 60.11482838 \t 0.7095593 \t 101.83344894 \t 0.70889957 \t 3.7679 \t 628.85157204\n",
      "199001 \t 60.09050112 \t 0.70970325 \t 101.8302446 \t 0.70893474 \t 3.76866 \t 638.07580686\n",
      "200001 \t 60.06917719 \t 0.70985692 \t 101.82482605 \t 0.7089797 \t 3.77059 \t 640.54982281\n",
      "201001 \t 60.07134571 \t 0.70982241 \t 101.81734688 \t 0.70901524 \t 3.76696 \t 643.17029428\n",
      "202001 \t 60.061887 \t 0.70995509 \t 101.81688782 \t 0.70901062 \t 3.76396 \t 638.45368862\n",
      "203001 \t 60.06498421 \t 0.70988502 \t 101.81642624 \t 0.70901481 \t 3.76753 \t 627.4793582\n",
      "204001 \t 60.07410877 \t 0.70982263 \t 101.81406502 \t 0.7090307 \t 3.76432 \t 635.7139585\n",
      "205001 \t 60.08093847 \t 0.7098134 \t 101.8168679 \t 0.70903544 \t 3.76683 \t 646.66666889\n",
      "206001 \t 60.06047321 \t 0.70995759 \t 101.82370056 \t 0.7090334 \t 3.76792 \t 641.39691186\n",
      "207001 \t 60.05398426 \t 0.70998853 \t 101.82904411 \t 0.70898981 \t 3.76509 \t 635.93376279\n",
      "208001 \t 60.04060572 \t 0.71014498 \t 101.82582741 \t 0.70902534 \t 3.77033 \t 637.17251372\n",
      "209001 \t 60.02429226 \t 0.71031938 \t 101.8152858 \t 0.70910789 \t 3.77327 \t 633.10617304\n",
      "210001 \t 59.99487787 \t 0.71053954 \t 101.79599052 \t 0.70919895 \t 3.77332 \t 635.08455062\n",
      "211001 \t 59.9815958 \t 0.71067436 \t 101.79687073 \t 0.7091635 \t 3.76517 \t 646.83618879\n",
      "212001 \t 59.9866378 \t 0.71066391 \t 101.79016335 \t 0.70918949 \t 3.76724 \t 640.37810946\n",
      "213001 \t 59.99380253 \t 0.71062622 \t 101.77806931 \t 0.7092201 \t 3.76731 \t 640.32628822\n",
      "214001 \t 60.01040762 \t 0.71053891 \t 101.77426254 \t 0.70923467 \t 3.76906 \t 631.79396534\n",
      "215001 \t 60.03057745 \t 0.71038913 \t 101.77454659 \t 0.70921589 \t 3.76639 \t 630.81699347\n",
      "216001 \t 60.01718117 \t 0.71040977 \t 101.76460999 \t 0.70925218 \t 3.76884 \t 639.02907681\n",
      "217001 \t 60.01526138 \t 0.71037726 \t 101.75820717 \t 0.70928534 \t 3.76891 \t 641.84137416\n",
      "218001 \t 60.03116493 \t 0.71019784 \t 101.75617317 \t 0.70929524 \t 3.76831 \t 639.16668177\n",
      "219001 \t 60.02023069 \t 0.71025051 \t 101.75347832 \t 0.70931412 \t 3.76881 \t 633.99412537\n",
      "220001 \t 60.02053291 \t 0.71022575 \t 101.75480759 \t 0.70931064 \t 3.7683 \t 632.29120803\n",
      "221001 \t 60.00340679 \t 0.71037106 \t 101.75681656 \t 0.70928128 \t 3.76851 \t 643.98464489\n",
      "222001 \t 60.00762711 \t 0.71038596 \t 101.75499184 \t 0.70927694 \t 3.76778 \t 640.86442828\n",
      "223001 \t 60.00412777 \t 0.71047842 \t 101.75269089 \t 0.70930921 \t 3.76522 \t 633.77751994\n",
      "224001 \t 60.00914419 \t 0.71043073 \t 101.76106178 \t 0.70927219 \t 3.76755 \t 633.29459047\n",
      "225001 \t 59.99753761 \t 0.71051644 \t 101.76220695 \t 0.70929602 \t 3.76655 \t 635.31647539\n",
      "226001 \t 60.00461479 \t 0.71036511 \t 101.76373833 \t 0.70931773 \t 3.76253 \t 635.76482725\n",
      "227001 \t 60.01650675 \t 0.71022667 \t 101.76345367 \t 0.70932845 \t 3.76253 \t 642.45095921\n",
      "228001 \t 60.01359623 \t 0.71019395 \t 101.77449471 \t 0.70928692 \t 3.76291 \t 640.05982423\n",
      "229001 \t 60.00399336 \t 0.71021623 \t 101.76996414 \t 0.70932175 \t 3.76359 \t 632.21949482\n",
      "230001 \t 60.00468689 \t 0.71012011 \t 101.76834961 \t 0.70930336 \t 3.76275 \t 632.04202867\n",
      "231001 \t 60.00720741 \t 0.71013011 \t 101.76486496 \t 0.70930846 \t 3.7627 \t 630.39564419\n",
      "232001 \t 60.00598616 \t 0.71015791 \t 101.76418671 \t 0.70932597 \t 3.76587 \t 646.02869272\n",
      "233001 \t 60.01384239 \t 0.71015031 \t 101.76112213 \t 0.70933526 \t 3.76341 \t 644.02573586\n",
      "234001 \t 60.0114427 \t 0.71023574 \t 101.76173027 \t 0.70932541 \t 3.76291 \t 643.15309715\n",
      "235001 \t 60.00870171 \t 0.71027767 \t 101.75955963 \t 0.70930971 \t 3.76464 \t 635.55312276\n",
      "236001 \t 59.98582872 \t 0.71046188 \t 101.7572731 \t 0.70930578 \t 3.76227 \t 629.6808064\n",
      "237001 \t 59.97992393 \t 0.71057191 \t 101.74963501 \t 0.70933767 \t 3.76013 \t 641.3163712\n",
      "238001 \t 59.97171543 \t 0.71064285 \t 101.7493438 \t 0.70935318 \t 3.76232 \t 649.62735748\n",
      "239001 \t 59.98530037 \t 0.71050969 \t 101.75899483 \t 0.70930675 \t 3.76458 \t 641.88268876\n",
      "240001 \t 59.9928784 \t 0.71035357 \t 101.75951439 \t 0.709279 \t 3.76087 \t 630.46438622\n",
      "241001 \t 59.99452553 \t 0.71028645 \t 101.75491669 \t 0.70927739 \t 3.76587 \t 627.27590275\n",
      "242001 \t 60.00075872 \t 0.71022508 \t 101.74033218 \t 0.70936288 \t 3.7621 \t 640.27546906\n",
      "243001 \t 60.01032122 \t 0.71016072 \t 101.73619576 \t 0.70939209 \t 3.76194 \t 648.23615861\n",
      "244001 \t 60.01967627 \t 0.71008809 \t 101.73604637 \t 0.7094009 \t 3.76587 \t 640.0119257\n",
      "245001 \t 60.00730951 \t 0.71014909 \t 101.73962135 \t 0.70938622 \t 3.76313 \t 635.00015855\n",
      "246001 \t 60.00858501 \t 0.71011887 \t 101.74909561 \t 0.709348 \t 3.76161 \t 631.13857365\n",
      "247001 \t 60.012672 \t 0.71006108 \t 101.75007942 \t 0.70933677 \t 3.76493 \t 643.04553485\n",
      "248001 \t 60.00927025 \t 0.71008515 \t 101.74198257 \t 0.70933772 \t 3.76616 \t 649.49420166\n",
      "249001 \t 60.01164053 \t 0.70998789 \t 101.73939384 \t 0.70934753 \t 3.76564 \t 644.52066636\n",
      "250001 \t 59.99708307 \t 0.71008133 \t 101.734366 \t 0.70938591 \t 3.76593 \t 641.71516228\n",
      "251001 \t 60.00195107 \t 0.71005521 \t 101.7273658 \t 0.70941641 \t 3.76873 \t 633.45712328\n",
      "252001 \t 59.99366291 \t 0.71013968 \t 101.71975838 \t 0.70946717 \t 3.7698 \t 633.05626464\n",
      "253001 \t 59.99831326 \t 0.71015897 \t 101.71936272 \t 0.70947783 \t 3.76576 \t 640.02853274\n",
      "254001 \t 60.00517227 \t 0.71010946 \t 101.71180351 \t 0.709529 \t 3.76624 \t 651.98773241\n",
      "255001 \t 60.00968624 \t 0.71007527 \t 101.7040934 \t 0.70954443 \t 3.76294 \t 633.87888145\n",
      "256001 \t 60.00679672 \t 0.71009272 \t 101.70142441 \t 0.70954236 \t 3.7623 \t 641.04162645\n",
      "257001 \t 60.00784859 \t 0.71009543 \t 101.69943672 \t 0.70955529 \t 3.764 \t 632.20201349\n",
      "258001 \t 60.01384678 \t 0.71009551 \t 101.69297905 \t 0.70955887 \t 3.76749 \t 637.91770196\n",
      "259001 \t 60.0182597 \t 0.71013674 \t 101.70206429 \t 0.70951451 \t 3.76605 \t 637.13378358\n",
      "260001 \t 60.0258087 \t 0.71007938 \t 101.70858368 \t 0.7094667 \t 3.76674 \t 642.42418432\n",
      "261001 \t 60.01688478 \t 0.71017536 \t 101.70942352 \t 0.70946641 \t 3.76449 \t 636.32459807\n",
      "262001 \t 60.01368137 \t 0.71023386 \t 101.72051437 \t 0.70941501 \t 3.7639 \t 623.04276252\n",
      "263001 \t 59.99254137 \t 0.7104115 \t 101.71808922 \t 0.70942701 \t 3.7662 \t 644.79670477\n",
      "264001 \t 59.99204973 \t 0.71042481 \t 101.7206913 \t 0.7094137 \t 3.76686 \t 642.63697267\n",
      "265001 \t 59.99454811 \t 0.7104513 \t 101.73078453 \t 0.70936749 \t 3.76814 \t 640.60351014\n",
      "266001 \t 59.98099438 \t 0.71065219 \t 101.73468269 \t 0.70939066 \t 3.76606 \t 634.43608689\n",
      "267001 \t 59.96391878 \t 0.71077088 \t 101.73229645 \t 0.70942683 \t 3.76639 \t 634.46927738\n",
      "268001 \t 59.98074366 \t 0.71056023 \t 101.72746048 \t 0.7094405 \t 3.76885 \t 647.58072495\n",
      "269001 \t 60.00024172 \t 0.71036787 \t 101.71451714 \t 0.70944646 \t 3.76833 \t 639.04722214\n",
      "270001 \t 59.99760328 \t 0.7102959 \t 101.6984864 \t 0.70950613 \t 3.76941 \t 650.43784022\n",
      "271001 \t 59.98245174 \t 0.7103744 \t 101.68884842 \t 0.70955394 \t 3.76938 \t 637.19759345\n",
      "272001 \t 59.9825881 \t 0.71040163 \t 101.68853378 \t 0.70952583 \t 3.77184 \t 633.1472578\n",
      "273001 \t 59.9713175 \t 0.7105423 \t 101.67557549 \t 0.70955539 \t 3.76722 \t 639.15206695\n",
      "274001 \t 59.987675 \t 0.7104246 \t 101.67476372 \t 0.70953417 \t 3.76789 \t 640.42123532\n",
      "275001 \t 60.00189169 \t 0.71022737 \t 101.67458916 \t 0.70957191 \t 3.76547 \t 645.03801584\n",
      "276001 \t 60.00816244 \t 0.71022809 \t 101.67549644 \t 0.70959032 \t 3.76529 \t 645.55696058\n",
      "277001 \t 59.99887009 \t 0.71033681 \t 101.68389862 \t 0.70955666 \t 3.76367 \t 640.78900051\n",
      "278001 \t 60.00847176 \t 0.71034302 \t 101.68787788 \t 0.70953361 \t 3.76479 \t 630.14667034\n",
      "279001 \t 59.99637251 \t 0.71043622 \t 101.6796814 \t 0.70955457 \t 3.76704 \t 635.59094429\n",
      "280001 \t 59.99950747 \t 0.71048443 \t 101.68368935 \t 0.7095407 \t 3.76739 \t 645.68596363\n",
      "281001 \t 60.00016634 \t 0.71045848 \t 101.6746254 \t 0.70954502 \t 3.76801 \t 645.40269852\n",
      "282001 \t 60.00568386 \t 0.71037187 \t 101.68186295 \t 0.7094949 \t 3.76884 \t 636.65226817\n",
      "283001 \t 60.00936432 \t 0.71031321 \t 101.68805031 \t 0.70946206 \t 3.77006 \t 626.26338816\n",
      "284001 \t 60.02354338 \t 0.7102302 \t 101.68904449 \t 0.70946766 \t 3.77034 \t 637.07805991\n",
      "285001 \t 60.00763996 \t 0.71035808 \t 101.68012672 \t 0.7095293 \t 3.769 \t 637.61421108\n",
      "286001 \t 60.01597324 \t 0.71032892 \t 101.68038567 \t 0.70955436 \t 3.76669 \t 645.62709188\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "287001 \t 60.02030403 \t 0.71028445 \t 101.6856691 \t 0.70954339 \t 3.76861 \t 639.02747869\n",
      "288001 \t 60.01919922 \t 0.71032221 \t 101.68390106 \t 0.70956156 \t 3.767 \t 631.31878495\n",
      "289001 \t 60.01338963 \t 0.71029423 \t 101.67665787 \t 0.70957582 \t 3.76771 \t 637.43450141\n",
      "290001 \t 60.01219015 \t 0.71031796 \t 101.67185783 \t 0.7096061 \t 3.7678 \t 645.11453152\n",
      "291001 \t 59.99674229 \t 0.71043551 \t 101.66080132 \t 0.70965529 \t 3.76893 \t 637.62152863\n",
      "292001 \t 60.00278276 \t 0.71029436 \t 101.65573723 \t 0.70969187 \t 3.7688 \t 640.55098367\n",
      "293001 \t 60.00891653 \t 0.71026776 \t 101.65192825 \t 0.70972322 \t 3.76679 \t 628.22982597\n",
      "294001 \t 59.99982479 \t 0.71031217 \t 101.65667297 \t 0.7096906 \t 3.76592 \t 632.66432619\n",
      "295001 \t 59.98406664 \t 0.71038894 \t 101.65694817 \t 0.70970389 \t 3.76694 \t 635.00568962\n",
      "296001 \t 59.98406174 \t 0.71035673 \t 101.65791389 \t 0.7097221 \t 3.76955 \t 645.42015386\n",
      "297001 \t 59.96171362 \t 0.71058993 \t 101.66064262 \t 0.7097194 \t 3.76849 \t 644.6555016\n",
      "298001 \t 59.96223572 \t 0.71062003 \t 101.65895081 \t 0.70971487 \t 3.76591 \t 633.15469861\n",
      "299001 \t 59.95785223 \t 0.71069802 \t 101.66215042 \t 0.70970372 \t 3.76413 \t 627.98285532\n",
      "\n",
      "TRAINING DONE\n"
     ]
    }
   ],
   "source": [
    "with tf.Session(graph=graph) as session:\n",
    "    tf.global_variables_initializer().run()\n",
    "    print(\"Initialized\")\n",
    "    ptime = time.time()\n",
    "    tfconfig = tf.ConfigProto()\n",
    "    tfconfig.gpu_options.allow_growth=gpu_memory_growth\n",
    "    init_op = tf.global_variables_initializer()    \n",
    "    sess = tf.Session(config=tfconfig)\n",
    "    sess.run(init_op)\n",
    "    print ('iter\\t test_squ_loss\\t test_avg_prec\\t',\n",
    "           'train_squ_loss\\t train_avg_prec\\t regularization\\t time (in s)')\n",
    "    for step in range(num_steps):        \n",
    "        #print('Step',step)\n",
    "        #print('There are {} samples in the test set'.format(len(model_stats.test_ids)))\n",
    "        #print('There are {} samples in the train set'.format(len(model_stats.train_ids)))\n",
    "       \n",
    "        xmb, ymb = get_training_batch(model_stats, batch_size, window, \n",
    "                                      pitch_transforms, jitter)\n",
    "        \n",
    "        # every 1000-nd step save the results           \n",
    "        if step%control_step==0:\n",
    "            # training step \n",
    "            # to check output the loss (save the loss but you cannot compare it with the mse_train)\n",
    "            _, np_loss, np_global_step, np_lr, reg_np = \\\n",
    "            sess.run([training_op, loss, global_step, lr, reg], \n",
    "                                     feed_dict={xb: xmb, yb: ymb})\n",
    "            \n",
    "            # A) run model on test set\n",
    "                        \n",
    "            # 1. get the test set\n",
    "            X,Y = get_test_sample(model_stats.test_ids, window, 1000)\n",
    "            Y = Y[:,base_note:base_note+mm] # in general without effect\n",
    "            # shape Xtest: (1000*len(test_ids), 1, 16384, 1)\n",
    "            # shape Ytest: (1000*len(test_ids), 128)\n",
    "            # extended test-set: 11 samples\n",
    "\n",
    "            \n",
    "            # 2. prediction of output and calculation of\n",
    "            # - MSE\n",
    "            # - average precision\n",
    "            Y_p, mse_test = predict_direct_model(sess, X, Y)\n",
    "            avp_test = average_precision_score(Y.flatten(), Y_p.flatten())\n",
    "            \n",
    "            del X,Y,Y_p\n",
    "            \n",
    "            # B) run model on training set\n",
    "            \n",
    "            # 1. get training sample\n",
    "            X,Y = get_training_sample(window)\n",
    "            Y = Y[:,base_note:base_note+mm] # in general without effect\n",
    "            # shape Xtrain: (15000, 1, 16384, 1)\n",
    "            # shape Ytrain: (15000, 128)\n",
    "            \n",
    "            # 2. prediction of output and calculation of\n",
    "            # - MSE\n",
    "            # - average precision\n",
    "            Y_p, mse_train = predict_direct_model(sess, X, Y)\n",
    "            avp_train = average_precision_score(Y.flatten(), Y_p.flatten())\n",
    "            \n",
    "            del X,Y,Y_p\n",
    "            \n",
    "            # c) Print the temporary results\n",
    "            print (np_global_step,'\\t', round(mse_test,8),\n",
    "                    '\\t', round(avp_test,8),\n",
    "                    '\\t', round(mse_train,8),\n",
    "                    '\\t', round(avp_train,8),\n",
    "                    '\\t', round(reg_np,8),\n",
    "                    '\\t', round(time.time() - ptime,8))\n",
    "            \n",
    "            #D) Save the temporary results\n",
    "            model_stats.stats['iter'][2].append(np_global_step)\n",
    "            # protocol learning rate\n",
    "            model_stats.stats['lr'][2].append(np_lr)\n",
    "            # protocol runtime\n",
    "            model_stats.stats['time'][2].append(time.time() - ptime)\n",
    "            # protocol weights and its norms\n",
    "            weight_norms = dict()\n",
    "            for name, w in model_stats.weights.items():\n",
    "                weight_norms[name] = np.mean(np.linalg.norm(w.eval(session=sess),axis=0))\n",
    "            for name, norm in weight_norms.items():\n",
    "                model_stats.stats['n'+name][2].append(norm)\n",
    "            # protocol training errors     \n",
    "            track = []\n",
    "            model_stats.stats['mse_train_1'][2].append(np_loss)\n",
    "            model_stats.stats['mse_train'][2].append(mse_train)\n",
    "            track += [tf.summary.scalar('mse_train', mse_train)]\n",
    "            model_stats.stats['avp_train'][2].append(avp_train)\n",
    "            track += [tf.summary.scalar('avp_train', avp_train)]\n",
    "            # protocol test errors\n",
    "            model_stats.stats['mse_test'][2].append(mse_test)\n",
    "            track += [tf.summary.scalar('mse_test', mse_test)]\n",
    "            model_stats.stats['avp_test'][2].append(avp_test)\n",
    "            track += [tf.summary.scalar('avp_test', avp_test)]\n",
    "            ptime = time.time()\n",
    "            \n",
    "                   \n",
    "        else:\n",
    "            #training step \n",
    "            sess.run([training_op], feed_dict={xb: xmb, yb: ymb})\n",
    "        \n",
    "        if step/num_steps==0.5 or (step+1)/num_steps==1:\n",
    "            saver.save(sess, 'models_'+notebook_reference+'/model', global_step=step)            \n",
    "        \n",
    "    print ('\\nTRAINING DONE')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'models_1903143/model-299999'"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saver.save(sess, 'models_'+notebook_reference+'/model', global_step=step) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 7. Post Processing\n",
    "\n",
    "Define functions to save the model_results diccionary to a file and load it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# in case diccionary should be printed\n",
    "def print_dict(dictionary):\n",
    "    for key, value in  dictionary.items():\n",
    "        print (key)\n",
    "        print (value)\n",
    "#print_dict(model_stats.stats)\n",
    "    \n",
    "#save diccionary\n",
    "def save_obj(obj, name ):\n",
    "    with open(name + '.pkl', 'wb') as f:\n",
    "        pickle.dump(obj, f, pickle.HIGHEST_PROTOCOL)\n",
    "\n",
    "# load diccionary\n",
    "def load_obj(name ):\n",
    "    with open(name + '.pkl', 'rb') as f:\n",
    "        return pickle.load(f)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 7.1 Plots"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# save diccionary to have a look at outputs at other times\n",
    "save_obj(model_stats.stats, 'results/'+notebook_reference+'_stats')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "results_stats = load_obj('results/'+notebook_reference+'_stats')\n",
    "\n",
    "# size of plots\n",
    "width = 20\n",
    "height = 13\n",
    "fontsize = 24\n",
    "control_step = 1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7f7614066d68>"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f761407d748>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "iters_res = results_stats['iter'][2]\n",
    "time_res = results_stats['time'][2]\n",
    "lr_res = results_stats['lr'][2]\n",
    "\n",
    "plt.figure(figsize=(width, height/2))\n",
    "plt.rcParams['xtick.labelsize']=fontsize\n",
    "plt.rcParams['ytick.labelsize']=fontsize\n",
    "#plot time needed for 'control_step' steps\n",
    "\"\"\"\n",
    "plt.subplot(2, 1, 1)\n",
    "plt.title('Runtime and Learning Rate')\n",
    "plt.plot(iters_res[1:], time_res[1:], c=\"darkblue\")\n",
    "plt.ylabel('Runtime of last '+str(control_step)+' steps (sec)')\n",
    "plt.ylim( (0.9*np.min(time_res[1:]), 1.1*np.max(time_res[1:])))\n",
    "\n",
    "plt.subplot(2, 1, 2)\n",
    "\"\"\"\n",
    "plt.subplot(1, 1, 1)\n",
    "plt.plot(iters_res[1:], lr_res[1:], c=\"darkblue\")\n",
    "plt.ylabel('Learning rate',fontsize=fontsize)\n",
    "plt.xlabel('Iteration',fontsize=fontsize)\n",
    "\n",
    "#plt.savefig('results/plots/'+notebook_reference+'_runtime_lr.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7f761417a048>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7614128400>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mse_train_res = results_stats['mse_train'][2]\n",
    "mse_test_res = results_stats['mse_test'][2]\n",
    "\n",
    "avp_train_res = results_stats['avp_train'][2]\n",
    "avp_test_res = results_stats['avp_test'][2]\n",
    "\n",
    "plt.figure(figsize=(width, height))\n",
    "\n",
    "#plot mean squared error\n",
    "plt.subplot(2, 1, 1)\n",
    "#plt.title('Error and Precision',fontsize=fontsize)\n",
    "plt.plot(iters_res, mse_train_res, label='Training', c=\"c\")\n",
    "plt.plot(iters_res, mse_test_res, label='Test', c=\"darkblue\")\n",
    "plt.legend(fontsize=fontsize)\n",
    "plt.ylabel('Mean Square Error',fontsize=fontsize)\n",
    "\n",
    "#plot average precision\n",
    "plt.subplot(2, 1, 2)\n",
    "plt.plot(iters_res, avp_train_res, label='Training', c=\"c\")\n",
    "plt.plot(iters_res, avp_test_res, label='Test', c=\"darkblue\")\n",
    "plt.legend(fontsize=fontsize)\n",
    "plt.ylabel('Average Precision',fontsize=fontsize)\n",
    "plt.xlabel('Iteration',fontsize=fontsize)\n",
    "\n",
    "#plt.savefig('results/plots/'+notebook_reference+'_error_prec.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7f761463f438>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7614772ac8>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nw = results_stats['nw'][2]\n",
    "nw2 = results_stats['nw2'][2]\n",
    "nbeta = results_stats['nbeta'][2]\n",
    "\n",
    "plt.figure(figsize=(width, height))\n",
    "\n",
    "plt.subplot(2, 1, 1)\n",
    "#plt.title('Norm of the Weights of Hidden and Output Layer',fontsize=fontsize)\n",
    "plt.plot(iters_res, nw, label='2nd layer', c=\"darkblue\")\n",
    "plt.plot(iters_res, nw2, label='3rd layer', c=\"c\")\n",
    "\n",
    "plt.legend(fontsize=fontsize)\n",
    "plt.ylabel('l2-Norm',fontsize=fontsize)\n",
    "\n",
    "plt.subplot(2, 1, 2)\n",
    "\n",
    "plt.plot(iters_res, nbeta, label='Output layer', c=\"darkblue\")\n",
    "plt.legend(fontsize=fontsize)\n",
    "plt.ylabel('l2-Norm',fontsize=fontsize)\n",
    "plt.xlabel('Iteration',fontsize=fontsize)\n",
    "\n",
    "#plt.savefig('results/plots/'+notebook_reference+'_weight_norms.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 7.2 Mirex Statistics\n",
    "\n",
    "Statistics that are necesarry for evaluation. The autuor used a threshold of 0.4 for calculation of error etc..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "def mirex_statistics(Y, Y_p, threshold=.5, base_note=0, mm=128):\n",
    "    \n",
    "    avp = average_precision_score(Y.flatten(), Y_p.flatten())\n",
    "\n",
    "    # we want zeros and ones\n",
    "    Yhat_p = Y_p>threshold\n",
    "    \n",
    "    # obtain frequencies\n",
    "    Yp_list = []\n",
    "    Y_list = []\n",
    "    \n",
    "    # for every sample\n",
    "    for i in range(len(Yhat_p)):\n",
    "        fhat = []\n",
    "        f = []\n",
    "        \n",
    "        for note in range(mm):\n",
    "            if Yhat_p[i][note] == 1:\n",
    "                fhat.append(440.*2**(((note+base_note) - 69.)/12.))\n",
    "            if Y[i][note] == 1:\n",
    "                f.append(440.*2**(((note+base_note) - 69.)/12.))\n",
    "\n",
    "        Yp_list.append(np.array(fhat))\n",
    "        Y_list.append(np.array(f))\n",
    "\n",
    "    # Precision, Recall, Accuracy, Substitution, Miss, False Alarm, and Total Error scores \n",
    "    # based both on raw frequency values and values mapped to a single octave (chroma)\n",
    "    \n",
    "    P,R,Acc,Esub,Emiss,Efa,Etot,cP,cR,cAcc,cEsub,cEmiss,cEfa,cEtot = \\\n",
    "    mir_eval.multipitch.metrics(np.arange(len(Y_list))/100., Y_list,\n",
    "                                np.arange(len(Yp_list))/100., Yp_list)\n",
    "\n",
    "\n",
    "    print('{:.2f}\\t{:.3f}\\t{:.3f}\\t{:.3f}\\t{:.3f}\\t{:.3f}\\t{:.3f}\\t{:.3f}\\t{:.3f}'.format(\n",
    "        threshold,100*avp,100*P,100*R,Acc,Etot,Esub,Emiss,Efa))\n",
    "\n",
    "    return avp,P,R,Acc,Etot\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from models_1903143/model-299999\n",
      "Model restored.\n"
     ]
    }
   ],
   "source": [
    "tf.reset_default_graph()\n",
    "\n",
    "num_steps = 300000\n",
    "\n",
    "# Add ops to save and restore all the variables.\n",
    "saver = tf.train.Saver({'wavg': wavg, 'w2avg': w2avg, 'betaavg': betaavg})\n",
    "\n",
    "# Later, launch the model, use the saver to restore variables from disk, and\n",
    "# do some work with the model.\n",
    "with tf.Session(graph=graph) as sess:\n",
    "    # Restore variables from disk.\n",
    "    saver.restore(sess, 'models_'+notebook_reference+'/model-'+str(num_steps-1))\n",
    "    print(\"Model restored.\")\n",
    "    # Check the values of the variables\n",
    "\n",
    "    # A) run model on test set\n",
    "\n",
    "    # 1. get the test set\n",
    "    X, Y = get_test_sample(model_stats.test_ids, window, 1000)\n",
    "    Y = Y[:,base_note:base_note+mm] # in general without effect\n",
    "    # shape Xtest: (1000*len(test_ids), 1, 16384, 1)\n",
    "    # shape Ytest: (1000*len(test_ids), 128)\n",
    "    # extended test-set: 11 samples\n",
    "\n",
    "\n",
    "    # 2. prediction of output and calculation of\n",
    "    # - MSE\n",
    "    # - average precision\n",
    "    Y_p, mse_test = predict_direct_model(sess, X, Y)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With the output of the test sample predicted calculate the mirex statistics:\n",
    "\n",
    "Average Precision (not from mireval), Precision, Recall, Accuracy, Total Error, Substitution, Miss, False Alarm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Thr\tAvgP\tP\tR\tAcc\tETot\tESub\tEmiss\tEfa\n",
      "0.30\t71.066\t63.283\t74.288\t0.519\t0.554\t0.135\t0.123\t0.296\n",
      "0.35\t71.066\t66.314\t70.266\t0.518\t0.525\t0.130\t0.168\t0.227\n",
      "0.40\t71.066\t68.924\t66.589\t0.512\t0.512\t0.122\t0.212\t0.178\n",
      "0.45\t71.066\t71.135\t62.270\t0.497\t0.517\t0.113\t0.264\t0.139\n",
      "0.50\t71.066\t73.317\t57.865\t0.478\t0.530\t0.102\t0.319\t0.109\n"
     ]
    }
   ],
   "source": [
    "print('Thr\\tAvgP\\tP\\tR\\tAcc\\tETot\\tESub\\tEmiss\\tEfa')\n",
    "for th in [.3, .35,.4, .45, .5]:\n",
    "    #print(\"When using threshold\",th)\n",
    "    avp,P,R,Acc,Etot = mirex_statistics(Y, Y_p, th)"
   ]
  },
  {
   "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}