DATA-TESTER-Modular-Decimated_Neurons.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# TESTING PHASE (ONLY LSM)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Tests the readouts weights using new simulations instead of the training data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# It's necessary to run:\n",
    "# ulimit -S -n 8192\n",
    "# before the ipython notebook in order to make sure the \"too many files open\" error is not going to appear during larger simulations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import sys\n",
    "from BEE import * # https://github.com/ricardodeazambuja/BEE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Makes possible to show the output from matplotlib inline\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "    \n",
    "# Makes the figures in the PNG format:\n",
    "# For more information see %config InlineBackend\n",
    "%config InlineBackend.figure_formats=set([u'png'])\n",
    "\n",
    "# plt.rcParams['figure.figsize'] = 5, 10\n",
    "\n",
    "import numpy\n",
    "\n",
    "import save_load_file as slf # https://github.com/ricardodeazambuja/Python-UTILS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#\n",
    "# Controls if the results are saved to a file\n",
    "#\n",
    "\n",
    "save2file = False"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Initialization and functions necessary to the 2 DOF arm simulations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sim_set = \"experiment_0001\" # basically is the name of the folder where the data is read/saved\n",
    "base_dir = \"simulation_data_00003\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# This data must follow what was used to generate the trajectories\n",
    "# or the arm model will behave differently!\n",
    "\n",
    "# Total time spent during the movement (in seconds)\n",
    "MT = 2.0\n",
    "\n",
    "# Simulation time step (in seconds)\n",
    "time_step = 2/1000.0\n",
    "\n",
    "total_steps = int(MT/time_step)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# S0, S1, E1 and W1 are the Baxter's joint names\n",
    "# Four joints => four input variables\n",
    "Number_of_input_variables=4 # defines how the liquid is going to be divided to receive the inputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#\n",
    "# Number of neurons at each input layer (defines the INPUT resolution of the system)\n",
    "#\n",
    "# >>>>>>>>> It MUST be smaller or equal to (NofN/Number_of_input_variables) <<<<<<<<<<<\n",
    "#\n",
    "Ninput = 150"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "original_joints=numpy.zeros((0,4)) #creates an empty array with the right shape...\n",
    "\n",
    "temp=numpy.load(base_dir+\"/\"+sim_set+\"/XY_movement_square.npy\")\n",
    "original_joints=numpy.concatenate((original_joints,temp[:1000]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA2gAAANYCAYAAACxbM6QAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDNAkSZIkqUH8f2TClosUMRQTAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10ee91650>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,10))\n",
    "plt.plot(original_joints)\n",
    "plt.title(\"Original joint curves\")\n",
    "plt.ylabel(\"angle (rad)\")\n",
    "plt.xlabel(\"step\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-0.59881723, -0.64028174,  2.04201794,  0.20936455])"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "joints=numpy.copy(original_joints)\n",
    "joints[0,:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# In order to ONLY verify saved results, jump directly to the end of the notebook\n",
    "*[Results visualisation](#plot)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Configure the liquid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Liquid 0: [8151 4134 3840 4248 1353 7718]\n",
      "Liquid 1: [5285 1067 2851  158  672 9688]\n",
      "Liquid 2: [2981 5505 7474 1449 3473 9496]\n",
      "Liquid 3: [4926 7048 1558 1884 4054 8442]\n",
      "Liquid 4: [5165 5248 3543 5505  104 1396]\n"
     ]
    }
   ],
   "source": [
    "liqseeds = slf.load_from_file_gz(\"./\"+base_dir+\"/\"+sim_set+\"/\"+\"liqseeds.gzpickle\")\n",
    "number_of_liquids = len(liqseeds)\n",
    "for i in range(number_of_liquids):\n",
    "    print \"Liquid %d:\" % i, liqseeds[i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def generates_connections(liqseeds,lsm_i,neuron_params):\n",
    "    \n",
    "    disconnected,membrane_rand,current_rand,noisy_current_rand,SpkLiq_vresets_rand,lbd_value,parameter_list=neuron_params\n",
    "\n",
    "    #\n",
    "    # Liquid's shape\n",
    "    #\n",
    "\n",
    "    # Liquid's shape\n",
    "    net_shape = numpy.array([20,5,6])\n",
    "\n",
    "    # Random seeds:\n",
    "    # // RANDOM-1: Noisy offset currents / resets\n",
    "    # // RANDOM-2: Selection of the inhibitory and excitatory neurons\n",
    "    # // RANDOM-3: Internal connections of the liquid\n",
    "    RANDOM_1=liqseeds[lsm_i][1]\n",
    "    RANDOM_2=liqseeds[lsm_i][2]\n",
    "    RANDOM_3=liqseeds[lsm_i][3]\n",
    "    #DEBUG!\n",
    "    # RANDOM_4=liqseeds[lsm_i][4]\n",
    "    # RANDOM_0=liqseeds[lsm_i][0]\n",
    "\n",
    "    input_noise_std = 1 # Standard deviation of the input gaussian noise\n",
    "\n",
    "    # Neuron parameters\n",
    "    taum=0.03 # membrane time constant\n",
    "    cm=3e-08 # membrane capacitance\n",
    "    taue=0.003 # excitatory synapse time constant\n",
    "    taui=0.006 # inhibitory synapse time constant\n",
    "\n",
    "\n",
    "    \n",
    "    input_gain=105. # gain used for the input neurons\n",
    "    w_SD = 3.0 # input neurons standard deviation\n",
    "\n",
    "\n",
    "    #\n",
    "    # END OF THE SETUP - everything that should change between liquids is expressed above...\n",
    "    #\n",
    "\n",
    "\n",
    "    #\n",
    "    # This first initialization of the simulator is only used to generate the liquid structure's data\n",
    "    #\n",
    "\n",
    "    # Resets the simulator\n",
    "    BEE_free()\n",
    "\n",
    "    tstp = time_step\n",
    "\n",
    "    seeds=numpy.array(numpy.random.randint(0,10000,5),dtype=numpy.uint32)\n",
    "    # // RANDOM-0: Membrane initial potentials\n",
    "    # // RANDOM-1: Noisy offset currents / resets\n",
    "    # // RANDOM-2: Selection of the inhibitory and excitatory neurons\n",
    "    # // RANDOM-3: Internal connections of the liquid\n",
    "    # // RANDOM-4: Noisy corrents\n",
    "    seeds[1] = RANDOM_1\n",
    "    seeds[2] = RANDOM_2\n",
    "    seeds[3] = RANDOM_3\n",
    "\n",
    "\n",
    "    if not BEE_initialized():\n",
    "        initialize_sim(my_net_shape = net_shape, \\\n",
    "                       my_lbd_value = lbd_value, \\\n",
    "                       my_seeds=seeds, \\\n",
    "                       SpkLiq_step=time_step,\\\n",
    "                       SpkLiq_threads_N = 5,\\\n",
    "                       SpkLiq_vresets = SpkLiq_vresets_rand,\\\n",
    "                       SpkLiq_membrane_rand=membrane_rand,\\\n",
    "                       SpkLiq_current_rand=current_rand,\\\n",
    "                       SpkLiq_taum=taum, SpkLiq_cm=cm, SpkLiq_taue=taue, SpkLiq_taui=taui, \\\n",
    "                       SpkLiq_noisy_current_rand=noisy_current_rand)\n",
    "\n",
    "\n",
    "    if parameter_list!=[]:\n",
    "        change_parameters(parameter_list)\n",
    "\n",
    "    if BEE_initialized() and (not BEE_connected()):\n",
    "        generate_connections()\n",
    "\n",
    "    #\n",
    "    # Cancels the connections, but keeps the parameters\n",
    "    #\n",
    "    if disconnected:\n",
    "        BEE_writes_SpkLiq_inh_connections(0)\n",
    "        BEE_writes_SpkLiq_exc_connections(0)\n",
    "\n",
    "    if BEE_initialized() and (not BEE_connected()):\n",
    "        process_connections()\n",
    "\n",
    "    NofN,NoINHN,NoEXCN,NoINHC,NoEXCC=output_stats(stats=0)\n",
    "\n",
    "\n",
    "    #\n",
    "    # Input converter\n",
    "    # \n",
    "\n",
    "    # This information is used to define the neurons receiving inputs (excitatory neurons):\n",
    "    inhibitory_index_L = output_inh_indices(NoINHN)\n",
    "    inhibitory_index_L.shape\n",
    "\n",
    "    inhibitory_index_L=[] # input signals through all neurons (not only excitatory ones)\n",
    "\n",
    "    #\n",
    "    # Input pattern generation\n",
    "    # \n",
    "\n",
    "    Number_of_neurons_lsm = NofN\n",
    "    Number_of_input_layers = Number_of_input_variables\n",
    "    Number_of_neurons_inputs = Ninput\n",
    "\n",
    "\n",
    "\n",
    "    # \n",
    "    # Gaussian generation\n",
    "    #\n",
    "\n",
    "    #\n",
    "    # Generates the connection pattern used. In the case below it's a gaussian \n",
    "    # distributed one based on the connection weights.\n",
    "    #\n",
    "\n",
    "    # In order to use SpiNNaker I should change from weight based to spike based (like in Oliveri's article).\n",
    "\n",
    "    inputs_indices = [] #inputs_indices[input_variable][ni]\n",
    "    inputs_weights = [] #inputs_weights[input_variable][ni]\n",
    "\n",
    "\n",
    "    def gaussian(lamb,n,nt):\n",
    "        '''\n",
    "        Generates a gaussian centered at 'n'\n",
    "        '''\n",
    "        return input_gain*numpy.exp(-((nt-n)**2)/(2*(lamb)**2)) #No energy normalization\n",
    "\n",
    "    weight_func = gaussian\n",
    "\n",
    "\n",
    "    liquid_input_layer_size = int(Number_of_neurons_lsm/float(Number_of_input_layers))\n",
    "\n",
    "    # Goes through the liquid to generate the proper connections, \n",
    "    # but dividing the liquid into the same number of input layers.\n",
    "    # Each input layer only connects to its own slice of the liquid.\n",
    "\n",
    "    centre_positions=numpy.linspace(0,liquid_input_layer_size-1,num=Number_of_neurons_inputs).astype(dtype=numpy.int)\n",
    "\n",
    "    for inp in range(Number_of_input_layers): #goes through the input layers\n",
    "        temp_i=[]\n",
    "        temp_w=[]        \n",
    "        for i,ic in zip(range(inp*Number_of_neurons_inputs,Number_of_neurons_inputs*(inp+1)),\\\n",
    "                        range(Number_of_neurons_inputs)): #goes through the input neurons\n",
    "            temp_i2 = []\n",
    "            temp_w2 = []\n",
    "            for j,ji in zip(range(inp*liquid_input_layer_size,liquid_input_layer_size*(inp+1)),\\\n",
    "                            range(liquid_input_layer_size)):\n",
    "                if j not in inhibitory_index_L:\n",
    "                    centre_position=centre_positions[ic]\n",
    "                    temp_weight=weight_func(w_SD,centre_position,ji)             \n",
    "                    if temp_weight>0: #guaranties only real spikes will be injected, saving cpu time\n",
    "                        temp_w2.append(temp_weight) # in nA\n",
    "                        temp_i2.append(j) # i=>input's neuron; j=>liquid's neuron                                        \n",
    "            temp_i.append(temp_i2)\n",
    "            temp_w.append(temp_w2)\n",
    "        inputs_indices.append(temp_i)\n",
    "        inputs_weights.append(temp_w)\n",
    "    \n",
    "    return (inputs_indices,inputs_weights)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# This is the simulation to test the readout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def loads_coeffs(lsm_i,disconnected=0):\n",
    "    \n",
    "    extra_name=str(lsm_i)+\"_Modular\"\n",
    "\n",
    "    filenames = [\"./\"+base_dir+\"/\"+sim_set+\"/coefficients_residues_\"+str(var_name)+\"_linear\"+extra_name+\".pickle\" \\\n",
    "                 for var_name in ['s0','s1','e1','w1']]\n",
    "\n",
    "    # Make sure the coefficients were created for the right size of network...\n",
    "    [c_s0,r_s0] = slf.load_from_file(filenames[0])\n",
    "    [c_s1,r_s1] = slf.load_from_file(filenames[1])\n",
    "    [c_e1,r_e1] = slf.load_from_file(filenames[2])\n",
    "    [c_w1,r_w1] = slf.load_from_file(filenames[3])\n",
    "\n",
    "    ranges_vector=slf.load_from_file(\"./\"+base_dir+\"/\"+sim_set+\"/ranges_0.pickle\")\n",
    "    bias_vector=slf.load_from_file(\"./\"+base_dir+\"/\"+sim_set+\"/bias_0.pickle\")\n",
    "    \n",
    "    return ((c_s0,c_s1,c_e1,c_w1,r_s0,r_s1,r_e1,r_w1),(bias_vector,ranges_vector))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from multiprocessing import Process, Pipe\n",
    "\n",
    "def simulator(args):\n",
    "        import sys\n",
    "        import BEE as BEE_python\n",
    "        \n",
    "        import membrane_lowpass_md\n",
    "        membrane_lowpass = membrane_lowpass_md.membrane_lowpass\n",
    "\n",
    "        import numpy\n",
    "        \n",
    "        #import os\n",
    "        #sys.stdout = open(str(os.getpid()) + \".out\", \"a\", buffering=0)\n",
    "        \n",
    "        pipes, sim_args = args\n",
    "        pipe_in,pipe_out = pipes\n",
    "        \n",
    "        # The seeds array must be a numpy.array, int32 with 5 terms.\n",
    "        pi,neuron_params,bias_vector,ranges_vector,inputs_indices,inputs_weights,coeffs,seeds, disconnected_neurons_indices = sim_args\n",
    "        \n",
    "        disconnected,membrane_rand,current_rand,noisy_current_rand,SpkLiq_vresets_rand,lbd_value,parameter_list = neuron_params\n",
    "        \n",
    "        c_s0,c_s1,c_e1,c_w1,r_s0,r_s1,r_e1,r_w1 = coeffs\n",
    "\n",
    "\n",
    "        #\n",
    "        # Liquid's shape\n",
    "        #\n",
    "        net_shape = numpy.array([20,5,6])\n",
    "        \n",
    "        # S0, S1, E1 and W1\n",
    "        Number_of_input_variables=4 # defines how the liquid is going to be divided to receive the inputs        \n",
    "\n",
    "        # Number of neurons at each input layer (defines the INPUT resolution of the system)\n",
    "        Ninput = 150\n",
    "\n",
    "        # Simulation time step (in seconds)\n",
    "        time_step = 2/1000.0\n",
    "\n",
    "        # This is the trajectory I want to test. It goes from 0 to 3\n",
    "        tji=0\n",
    "\n",
    "        number_of_joints = 4\n",
    "\n",
    "        # Creates the Ninput possible discrete values of angle for each joint\n",
    "        joints_values=numpy.empty((Ninput,number_of_joints))\n",
    "        for ji in range(number_of_joints):\n",
    "            joints_values[:,ji]=numpy.linspace(-1,1,Ninput)\n",
    "            \n",
    "\n",
    "        # Neuron parameters\n",
    "        taum=0.03 # membrane time constant\n",
    "        cm=3e-08 # membrane capacitance\n",
    "        taue=0.003 # excitatory synapse time constant\n",
    "        taui=0.006 # inhibitory synapse time constant    \n",
    "             \n",
    "\n",
    "        # Resets the simulator\n",
    "        BEE_python.BEE_free()\n",
    "        # It was only BEE_free...\n",
    "\n",
    "        # print \"Initializing the simulator...\"\n",
    "        if not BEE_python.BEE_initialized():\n",
    "            BEE_python.initialize_sim(my_net_shape = net_shape, \\\n",
    "                                      my_lbd_value = lbd_value, \\\n",
    "                                      my_seeds=seeds, \\\n",
    "                                      SpkLiq_step=time_step,\\\n",
    "                                      SpkLiq_threads_N = 5,\\\n",
    "                                      SpkLiq_vresets = SpkLiq_vresets_rand,\\\n",
    "                                      SpkLiq_membrane_rand=membrane_rand,\\\n",
    "                                      SpkLiq_current_rand=current_rand,\\\n",
    "                                      SpkLiq_taum=taum, SpkLiq_cm=cm, SpkLiq_taue=taue, SpkLiq_taui=taui, \\\n",
    "                                      SpkLiq_noisy_current_rand=noisy_current_rand)\n",
    "\n",
    "\n",
    "        if parameter_list!=[]:\n",
    "            BEE_python.change_parameters(parameter_list)\n",
    "\n",
    "        BEE_python.generate_connections()\n",
    "\n",
    "        #\n",
    "        # Cancels the connections, but keeps the parameters\n",
    "        #\n",
    "        if disconnected:\n",
    "            BEE_python.BEE_writes_SpkLiq_inh_connections(0)\n",
    "            BEE_python.BEE_writes_SpkLiq_exc_connections(0)\n",
    "\n",
    "\n",
    "        BEE_python.process_connections()\n",
    "\n",
    "        # print \"Verifies...\"\n",
    "        NofN,NoINHN,NoEXCN,NoINHC,NoEXCC=BEE_python.output_stats(stats=0)\n",
    "\n",
    "        connected_neurons_indices = numpy.ones(NofN,dtype=numpy.int32)\n",
    "        \n",
    "        if len(disconnected_neurons_indices) != 0:\n",
    "            connected_neurons_indices[disconnected_neurons_indices] = numpy.zeros(len(disconnected_neurons_indices),dtype=numpy.int32)\n",
    "            \n",
    "        \n",
    "        # Forces the simulator to calculate the values for all neurons (even the unconnected ones)\n",
    "        # This is important during the tests without connections\n",
    "        BEE_python.control_connected(connected_neurons_indices)\n",
    "\n",
    "        \n",
    "        # DEBUG #\n",
    "        # print BEE_python.output_connected(NofN)\n",
    "        # sys.exit(0)\n",
    "        # DEBUG #\n",
    "        \n",
    "        \n",
    "        # run_test\n",
    "        using_spikes = 0\n",
    "        membrane_time_constant = 30E-3\n",
    "        \n",
    "\n",
    "        #BEE_python.liquid_soft_reset(seeds) # Resets the simulator\n",
    "\n",
    "        # Initialize\n",
    "        ti=0\n",
    "\n",
    "        m_v=membrane_lowpass(NofN,membrane_time_constant) # Initialize the membrane\n",
    "\n",
    "        # Runs the simulation going step-by-step\n",
    "        # Stops at the yield keyword!\n",
    "        # I want to receive joint angles (the average) and return the result from the LSM\n",
    "        \n",
    "        pipe_out.send(\"LSM %d Ready!\"%pi)\n",
    "        \n",
    "        while True:\n",
    "            \n",
    "            input_angles = pipe_in.recv() # receives the joint values... blocks until something arrives\n",
    "            \n",
    "            if input_angles:\n",
    "                joint_s0,joint_s1,joint_e1,joint_w1 = input_angles\n",
    "            \n",
    "                s0_value = (joint_s0-bias_vector[0])/ranges_vector[0]\n",
    "                s1_value = (joint_s1-bias_vector[1])/ranges_vector[1]\n",
    "                e1_value = (joint_e1-bias_vector[2])/ranges_vector[2]\n",
    "                w1_value = (joint_w1-bias_vector[3])/ranges_vector[3]\n",
    "\n",
    "                input_spikes = [abs(joints_values[:,0]-s0_value).argmin(),\\\n",
    "                                abs(joints_values[:,1]-s1_value).argmin(),\\\n",
    "                                abs(joints_values[:,2]-e1_value).argmin(),\\\n",
    "                                abs(joints_values[:,3]-w1_value).argmin()]\n",
    "\n",
    "                exc_inputs = numpy.concatenate([numpy.array(inputs_indices[input_variable][ni]) for input_variable,ni in zip(range(len(input_spikes)),input_spikes)])\n",
    "                inh_inputs = []\n",
    "                exc_weights =  numpy.concatenate([numpy.array(inputs_weights[input_variable][ni]) for input_variable,ni in zip(range(len(input_spikes)),input_spikes)])\n",
    "                inh_weights = []\n",
    "\n",
    "                # Here the simulator is receiving the input spikes (and returning the output ones):\n",
    "                # r, current_t, received_spikes, OutputSpks, t = s.run_step(input_spikes) # Sends the spikes and reads the output\n",
    "                BEE_python.updates_sim(exc_inputs, inh_inputs, exc_weights*1E-9, inh_weights, len(exc_inputs), 0)\n",
    "\n",
    "            else: # the input was None - useful to compare with SpiNNaker\n",
    "                BEE_python.updates_sim([], [], [], [], 0, 0)\n",
    "\n",
    "            # reads LSM the output\n",
    "            OutputSpks = BEE_python.reads_spikes(NofN)\n",
    "            \n",
    "            # DEBUG #\n",
    "            #if len(OutputSpks):\n",
    "            #    print str(len(OutputSpks)) + \"-\",\n",
    "            #    sys.stdout.flush()\n",
    "            # DEBUG # \n",
    "\n",
    "            if using_spikes:\n",
    "                Output = numpy.zeros(NofN) \n",
    "                Output[OutputSpks]=1 # Using the output of the liquid directly\n",
    "            else:\n",
    "                m_v.process_spikes(OutputSpks,ti) # Send the spikes to the membrane filter\n",
    "                Output = m_v.check_values(ti) # Reads the values of the membranes\n",
    "        \n",
    "            joint_s0=(bias_vector[0]+(Output.dot(c_s0)+r_s0)*ranges_vector[0]) # Calculates the joint angles from the membrane filtered values\n",
    "            joint_s1=(bias_vector[1]+(Output.dot(c_s1)+r_s1)*ranges_vector[1]) # Calculates the joint angles from the membrane filtered values\n",
    "            joint_e1=(bias_vector[2]+(Output.dot(c_e1)+r_e1)*ranges_vector[2]) # Calculates the joint angles from the membrane filtered values\n",
    "            joint_w1=(bias_vector[3]+(Output.dot(c_w1)+r_w1)*ranges_vector[3]) # Calculates the joint angles from the membrane filtered values\n",
    "\n",
    "            ti+=time_step # advances the time\n",
    "            \n",
    "            pipe_out.send([[joint_s0,joint_s1,joint_e1,joint_w1],OutputSpks,Output,ti])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# New trials start here"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def kill_processes(new_sim,new_pipes_in,new_pipes_out):\n",
    "    #\n",
    "    # Kills the processes!\n",
    "    #\n",
    "    for i in range(len(new_sim)):\n",
    "        new_sim[i].terminate()\n",
    "        new_pipes_in[i][0].close()\n",
    "        new_pipes_in[i][1].close()        \n",
    "        new_pipes_out[i][0].close()\n",
    "        new_pipes_out[i][0].close()        \n",
    "        while new_sim[i].is_alive():\n",
    "            pass\n",
    "        # print \"Process %d is dead!\" % i\n",
    "\n",
    "    del new_sim\n",
    "    del new_pipes_in\n",
    "    del new_pipes_out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def initialize_LSM(total_parallel = 10, total_lsms = 0, lsm_i = 0, disconnected_neurons = []):\n",
    "    #\n",
    "    # Initialization of the LSM's\n",
    "    #\n",
    "    #total_parallel: total number of liquids to be used in PARALLEL (it is going to copy liquids if there aren't enough)\n",
    "    #\n",
    "    #total_lsms: len(liqseeds) # number of different LSMs available\n",
    "    #total_lsms: 0 Indicates one LSM is going to be used even if there are more available\n",
    "    #lsm_i: then it's number should be specified here (lsm_i), otherwise lsm_i is redefined inside this function\n",
    "\n",
    "\n",
    "    # Parameters paper Joshi/Maass\n",
    "    membrane_rand=[13.5E-3,14.9E-3]\n",
    "    current_rand=[13.5E-09, 14.5E-09]\n",
    "    noisy_current_rand=1E-9\n",
    "\n",
    "    factor=2*10.0\n",
    "\n",
    "    SpkLiq_vresets_rand=[13.8E-3,13.8E-3*(1+1/factor)] # the membrane reset values are drawn from a uniform distribution\n",
    "\n",
    "    lbd_value = 1.2\n",
    "\n",
    "    parameter_list=\\\n",
    "    [[[0.1  ,  0.32 ,  0.144,  0.06 ,  -47.  ,  0.8],\\\n",
    "      [0.4  ,  0.25 ,  0.7  ,  0.02 ,  -47  ,  0.8]],\\\n",
    "     [[0.2  ,  0.05 ,  0.125,  1.2  ,  150  ,  0.8],\\\n",
    "      [0.3  ,  0.5  ,  1.1  ,  0.05 ,  70  ,  1.5]]]\n",
    "\n",
    "\n",
    "    disconnected = 0\n",
    "\n",
    "    neuron_params = disconnected,\\\n",
    "                    membrane_rand,\\\n",
    "                    current_rand,\\\n",
    "                    noisy_current_rand,\\\n",
    "                    SpkLiq_vresets_rand,\\\n",
    "                    lbd_value,\\\n",
    "                    parameter_list\n",
    "\n",
    "\n",
    "\n",
    "    lsm_circ_arr = numpy.arange(total_lsms,dtype=numpy.int)\n",
    "\n",
    "    new_sim = []\n",
    "    new_pipes_in = []\n",
    "    new_pipes_out = []\n",
    "    joint_angles_mean = []\n",
    "    for i in range(total_parallel):\n",
    "        if total_lsms>0:\n",
    "            lsm_i = numpy.roll(lsm_circ_arr,-i)[0] # If you want to use always the same, do total_lsms=0 and set lsm_i.\n",
    "        new_pipes_in.append(Pipe(duplex=False))\n",
    "        new_pipes_out.append(Pipe(duplex=False))\n",
    "\n",
    "        coeffs, (bias_vector,ranges_vector) = loads_coeffs(lsm_i)\n",
    "\n",
    "        inputs_indices,inputs_weights = generates_connections(liqseeds,lsm_i,neuron_params)\n",
    "\n",
    "        seeds=numpy.array(numpy.random.randint(0,10000,5),dtype=numpy.uint32)\n",
    "        # // RANDOM-0: Membrane initial potentials\n",
    "        # // RANDOM-1: Noisy offset currents / resets <<<<<<\n",
    "        # // RANDOM-2: Selection of the inhibitory and excitatory neurons <<<<<<\n",
    "        # // RANDOM-3: Internal connections of the liquid <<<<<<\n",
    "        # // RANDOM-4: Noisy corrents    \n",
    "        #seeds[0] = liqseeds[lsm_i][0]\n",
    "        seeds[1] = liqseeds[lsm_i][1]\n",
    "        seeds[2] = liqseeds[lsm_i][2]\n",
    "        seeds[3] = liqseeds[lsm_i][3]\n",
    "        #seeds[4] = liqseeds[lsm_i][4]                \n",
    "        new_sim.append(Process(target=simulator, args=(((new_pipes_in[i][0], new_pipes_out[i][1]),\\\n",
    "                                                        (lsm_i,neuron_params,\\\n",
    "                                                         bias_vector,ranges_vector,\\\n",
    "                                                         inputs_indices,inputs_weights,\\\n",
    "                                                         coeffs,numpy.copy(seeds[:5]),disconnected_neurons[lsm_i])),)))\n",
    "\n",
    "    for i in range(total_parallel):\n",
    "        new_sim[i].daemon = True\n",
    "        new_sim[i].start()\n",
    "    \n",
    "    return new_sim, new_pipes_in, new_pipes_out, bias_vector, ranges_vector\n",
    "\n",
    "    # [Errno 24] Too many open files\n",
    "    # Try the command line tools:\n",
    "    # ulimit -a\n",
    "    # lsof | grep PIPE\n",
    "    # lsof | grep PIPE | grep -c Python\n",
    "    # What solved my problem was the use of:\n",
    "    # ulimit -S -n 8192 \n",
    "    # just before starting the ipython notebook..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def unlock_processes(new_sim,new_pipes_out):\n",
    "    #\n",
    "    # Unlock the processes\n",
    "    #\n",
    "    for i in range(len(new_sim)):\n",
    "        if new_sim[i].is_alive():\n",
    "            temp = new_pipes_out[i][0].recv() # unlocks the process\n",
    "            #print \"Process %d: %s\" % (i,temp)\n",
    "        else:\n",
    "            pass\n",
    "            #print \"%d - No processes to unlock!\" % i"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def run_trial(new_sim, new_pipes_in, new_pipes_out, simulation_type, trial_number, number_of_steps = 2000):\n",
    "    if simulation_type==\"parallel\":\n",
    "        #\n",
    "        # This version feeds back average values\n",
    "        #\n",
    "        # Each LSM receives the average of all the outputs.\n",
    "\n",
    "        joint_angles_individual = []\n",
    "        for i in range(len(new_sim)):\n",
    "            joint_angles_individual.append([])\n",
    "\n",
    "        for j in range(number_of_steps):\n",
    "            joint_angles = []    \n",
    "            for i in range(len(new_sim)):\n",
    "                [joint_s0,joint_s1,joint_e1,joint_w1] = joint_angles_mean[-1]\n",
    "                new_pipes_in[i][1].send([joint_s0,joint_s1,joint_e1,joint_w1])\n",
    "                ([joint_s0,joint_s1,joint_e1,joint_w1],OutputSpks,_,ti) = new_pipes_out[i][0].recv()\n",
    "                joint_angles.append([joint_s0,joint_s1,joint_e1,joint_w1])\n",
    "                joint_angles_individual[i].append([joint_s0,joint_s1,joint_e1,joint_w1])\n",
    "\n",
    "            joint_angles_mean.append(numpy.array(joint_angles).mean(axis=0))\n",
    "\n",
    "    return joint_angles_mean,joint_angles_individual"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Generation of the testing data using different configurations\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1h 14min 9s, sys: 6min 31s, total: 1h 20min 41s\n",
      "Wall time: 3h 43min 26s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "#\n",
    "# USING ALL THE AVAILABLE LIQUIDS\n",
    "#\n",
    "\n",
    "NofN = 600\n",
    "\n",
    "number_of_trials = 100\n",
    "\n",
    "temp = numpy.arange(NofN)\n",
    "\n",
    "for number_of_decimated_neurons in numpy.arange(11)*6:\n",
    "\n",
    "    extra_name = \"_Modular_decNeurons_\"+str(number_of_decimated_neurons)+\"_\"\n",
    "\n",
    "    simulation_type = \"parallel\"\n",
    "\n",
    "    parallel_lsms = 5\n",
    "\n",
    "    for trial_number in range(number_of_trials):\n",
    "        #\n",
    "        # Sets the initial joint angles => joints[0,:]\n",
    "        #\n",
    "        joint_angles_mean = []\n",
    "        joint_angles_mean.append(joints[0,:])   \n",
    "        \n",
    "        disconnected_neurons = []\n",
    "        \n",
    "        # Sets the disconnected neurons (randomly)\n",
    "        for lsm_i in range(parallel_lsms):\n",
    "            numpy.random.shuffle(temp)\n",
    "            disconnected_neurons.append(numpy.copy(temp[:number_of_decimated_neurons]))\n",
    "            \n",
    "\n",
    "        new_sim,new_pipes_in,new_pipes_out,bias_vector,ranges_vector = initialize_LSM(total_parallel = parallel_lsms, \\\n",
    "                                                                                      total_lsms = len(liqseeds), lsm_i = 0,\\\n",
    "                                                                                     disconnected_neurons = disconnected_neurons)\n",
    "        unlock_processes(new_sim,new_pipes_out)\n",
    "        joint_angles_mean,joint_angles_individual = run_trial(new_sim,new_pipes_in,new_pipes_out,simulation_type,trial_number)\n",
    "        kill_processes(new_sim,new_pipes_in,new_pipes_out)\n",
    "\n",
    "        filename = \"./\"+base_dir+\"/\"+sim_set+\"/joint_angles_mean_\"+simulation_type+\"_ALL_\"+str(trial_number)+extra_name+\"_LSM.npy\"\n",
    "        if save2file:\n",
    "            numpy.save(filename,joint_angles_mean)\n",
    "\n",
    "        filename = \"./\"+base_dir+\"/\"+sim_set+\"/joint_angles_individual\"+simulation_type+\"_ALL_\"+str(trial_number)+extra_name+\"_LSM.npy\"\n",
    "        if save2file:\n",
    "            numpy.save(filename,joint_angles_individual)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0,  6, 12, 18, 24, 30, 36, 42, 48, 54, 60])"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numpy.arange(11)*6"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plotting the results using the generated files (above)\n",
    "<a id=\"plot\"></a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "extra_name = \"_Modular_decNeurons_0_\"\n",
    "simulation_type = \"parallel\"\n",
    "\n",
    "lsm_i = \"ALL\"\n",
    "\n",
    "number_of_trials = 100\n",
    "\n",
    "joint_angles_mean = []\n",
    "for ti in range(number_of_trials):\n",
    "    filename = \"./\"+base_dir+\"/\"+sim_set+\"/joint_angles_mean_\"+simulation_type+\"_\"+str(lsm_i)+\"_\"+str(ti)+extra_name+\"_LSM.npy\"\n",
    "    joint_angles_mean.append(numpy.load(filename))\n",
    "\n",
    "ranges_vector=slf.load_from_file(\"./\"+base_dir+\"/\"+sim_set+\"/ranges_0.pickle\")\n",
    "bias_vector=slf.load_from_file(\"./\"+base_dir+\"/\"+sim_set+\"/bias_0.pickle\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "simulation_results = numpy.array(joint_angles_mean)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "joint_names = [\"s0\",\"s1\",\"e1\",\"w1\"]\n",
    "plt.figure(figsize =(30,30))\n",
    "for ji in range(4):\n",
    "    plt.subplot(411+ji)\n",
    "    for si in simulation_results:\n",
    "        plt.plot(si[:,ji],'g-',linewidth=2,label=\"Calculated\")\n",
    "    plt.plot(joints[:,ji],'b-',linewidth=4,label=\"Original\")\n",
    "    plt.title(\"Joint \"+joint_names[ji])\n",
    "#     plt.xlim(0,len(t_mov)-2)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  },
  "widgets": {
   "state": {},
   "version": "1.1.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}