IA Notebook #4 - Programando una Red Neuronal desde Cero.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# IA Notebook #3 - Programando una Red Neuronal desde Cero\n",
    "El problema a resolver está planteado en el [este enlace](https://youtu.be/uwbHOpp9xkc?t=500), básicamente, lo que se quiere es separar en dos grupos de puntos un conjunto de puntos. Un problema de clasificación.\n",
    "\n",
    "Para ello, primero haremos nuestro datasets y luego la red neuronal como tal. ;3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import scipy as sc\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from sklearn.datasets import make_circles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#************Creamos el dataset************\n",
    "\n",
    "#Número de registos que tenemos en nuestros datos:\n",
    "n=500\n",
    "#Número de características que tenemos sobre cada uno de nuestros datos:\n",
    "p=2\n",
    "#Para este ejercicio que se está desarrollando a modo de ejemplo, solo trabajamos con dos características\n",
    "#para poder dibujarlas en un plano 2D, a mayor número de características, mayor las dimensiones.\n",
    "\n",
    "#Para consultar los parametros de la función hacemos:\n",
    "#make_circle?\n",
    "#Nuestro dataset estará conformada por una entrada \"x\" y una salida \"y\":\n",
    "x, y = make_circles(n_samples=n, noise=0.05, factor=0.5)\n",
    "\n",
    "# Aquí corregimos la variable \"y\" que viene a ser de la forma (500, )\n",
    "y = y[:, np.newaxis] #Y ahora lo convertimos en un vector como tal -> (500, 1)\n",
    "#print(y.shape) #Para verificarlo podemos ejecutar esto.\n",
    "\n",
    "#Para visualizar como van quedando los datos, dibujamos:\n",
    "#plt.scatter(x[:, 0], x[:, 1], c=\"orange\")\n",
    "#plt.scatter(x[:, 1], x[:, 0], c=\"red\") #Solo lo puse a modo de prueba :3\n",
    "#plt.show()\n",
    "\n",
    "#Pero continuando con el ejercicio: (Y haciendo la correción de la \"y\")\n",
    "#plt.scatter(x[y==0, 0], x[y==0, 1], c=\"red\") #Con esto le decimos que solo dibuje uno de los circulos\n",
    "plt.scatter(x[y[:, 0]==0, 0], x[y[:, 0]==0, 1], c=\"red\")\n",
    "#plt.scatter(x[y==1, 0], x[y==1, 1], c=\"orange\")\n",
    "plt.scatter(x[y[:, 0]==1, 0], x[y[:, 0]==1, 1], c=\"orange\")\n",
    "plt.axis(\"equal\") #Con esto le decimos que esten en la misma proporción ambos ejes\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "##### Otra vez, haciendo aquí unas pruebas:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#print(y[:, 0])\n",
    "#print(y)\n",
    "plt.scatter(x[:, 0], y+0.25, c=\"orange\")\n",
    "plt.scatter(x[:, 1], y-0.25, c=\"gray\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "## Continuando...\n",
    "Ahora si, a programar la red neuronal! :D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#***********Estructura de datos que contiene nuestros parámetros de la red neuronal***********\n",
    "#No es necesario heredar de la clase \"object\" porque ya lo hace automáticamente:\n",
    "class neural_layer(): #Clase de la capa de la red neuronal\n",
    "  #Aquí podría inicializar los atributos:\n",
    "  n_conn=n_neur=act_f=0.0\n",
    "  b=w=np.zeros(1) #Definiendo a \"b\" y \"w\" como matrices o vectores unitarios con valor 0\n",
    "  #Y podría usar este constructor:\n",
    "  def __init__(self, numConn, numNeur, funcAct): #El cual SI FUNCIONA!\n",
    "  #def __init__(self, n_conn, n_neur, act_f):\n",
    "    #n_conn: número de conexiones que vienen de una capa anterior.\n",
    "    #n_neur: número de neuronas en nuestra capa\n",
    "    #act_f: función de activación en nuestra capa\n",
    "\n",
    "    self.act_f = funcAct\n",
    "    #self.act_f = act_f #Pero hacer las cosas de esta manera es más sencillo\n",
    "\n",
    "    self.n_conn=numConn #Sin embargo, así creo que es más formal.\n",
    "    self.n_neur=numNeur #Y lo es, pero Python es tan versátil que ni es necesario hacerlo.\n",
    "    #Así que estás líneas solo fueron opcinoales, podría hacerlo mucho más directo.\n",
    "    #Claro, solo con Python, otros lenguajes son más estrictos.\n",
    "\n",
    "    #Vector de vallas:\n",
    "    self.b = np.random.rand(1, numNeur)*2 -1\n",
    "    #self.b = np.random.rand(1, n_neur)*2 -1 #Esto es solo por si acaso falla.\n",
    "    #rand varía de 0 a 1 por eso se multiplica y resta para trabajar con red normalizada y estandarizada.\n",
    "    #Ya que estos serán los parámetros con los que inicializará la red neuronal\n",
    "    #(No lo apunté pero...) Este sería el vector de pesos:\n",
    "    self.w = np.random.rand(numConn, numNeur)*2 -1\n",
    "    #self.w = np.random.rand(numConn, numNeur)*2 -1 #Esto es solo por si acaso falla."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Funciones de activación (Si uno quiere puede definir más para hacer las pruebas)\n",
    "sigmoide = (#Función sigmoide\n",
    "            lambda x: 1/(1+np.e**(-x)),\n",
    "            #Derivada de la función sigmoide\n",
    "            lambda x: x*(1-x))\n",
    "relu = lambda x: np.maximum(0, x)\n",
    "\n",
    "#Vector o matriz con 100 valores de -5 a 5\n",
    "_x=np.linspace(-5, 5, 100)\n",
    "\n",
    "#plt.plot(_x, relu(_x), c=\"red\")\n",
    "plt.plot(_x, sigmoide[0](_x), c=\"skyblue\")\n",
    "#plt.plot(_x, sigmoide[1](_x), c=\"blue\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Se podría crear capa por capa de manera manual:\n",
    "#Capa 0\n",
    "#l0 = neural_layer(p, 4, sigmoide)\n",
    "#El primer parámetro es \"p\" porque es la cantidad de características que tenemos sobre nuestros datos.\n",
    "#Capa 1\n",
    "#l1 = neural_layer(4, 8, sigmoide)\n",
    "#La cantidad es ahora 4 porque el número de neuronas de la anterior capa es igual\n",
    "#Y el número de conexiones debe ser mínimo a esa cantidad.\n",
    "\n",
    "#---------------------------------------------------------\n",
    "#O hacerlo iterativo y para ello:\n",
    "\n",
    "#Creamos este vector o lista o matriz, que definirá el número de neuronas por cada capa:\n",
    "#topologia = [p, 4, 8, 16, 8, 4, 1] #El número ha sido arbitrario para este caso.\n",
    "#Y se comentá aquí solo para mantener una estética.\n",
    "\n",
    "#No necesariamente todas las capas deben tener la misma función de activación\n",
    "#Para este ejemplo si, por simplicidad:\n",
    "def crear_redNeuronal(topology, activation_function): #O \"create_nn\" en inglés.\n",
    "  #red_neuronal -> rn o en inglés: neural_network -> nn\n",
    "  nn=[]\n",
    "  #Ahora un bucle para recorrer todos lo valores y crear la red:\n",
    "  for indiceCapa, capa in enumerate(topology[:-1]): #Con el -1 le digo que recorra todo menos el último.\n",
    "    nn.append(neural_layer(topology[indiceCapa], topology[indiceCapa+1], activation_function))\n",
    "  \n",
    "  return nn\n",
    "\n",
    "#Aquí lo colocamos más ordenado: (pero nuevamente lo movemos y ahora al siguiente segmento)\n",
    "#topologia = [p, 4, 8, 16, 8, 4, 1]\n",
    "#Y creamos red neuronal:\n",
    "#crear_redNeuronal(topologia, sigmoide)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Se define la topología:\n",
    "#topologia = [p, 4, 8, 16, 8, 4, 1]\n",
    "#Creamos la red neuronal:\n",
    "#red_neuronal=crear_redNeuronal(topologia, sigmoide) #red_neuronal -> rn\n",
    "#Todo esto anterior se usó para probar la función \"entrenar\".\n",
    "\n",
    "#Definimos dos funciones de coste (vamos a usar el error cuadrático medio)\n",
    "#La primera será la función como tal y la segunda será su derivada:\n",
    "i2_cost = (\n",
    "    #Error cuadrático medio:\n",
    "    lambda Ypredicha, Yreal: np.mean((Ypredicha-Yreal)**2), #Opera y luego cálcula la media.\n",
    "    #Derivada del error cuadrático medio:\n",
    "    lambda Ypredicha, Yreal: Ypredicha-Yreal\n",
    ")\n",
    "\n",
    "#Recordar que el \"lr\" es el largo del paso que damos en el \"Descenso del gradiente\"\n",
    "def entrenar(rn, _x, _y, func_Coste, lr=0.5, train=True):\n",
    "  #En este vector guardaremos esta información:\n",
    "  #out = [(z0, a0), (z1, a1), etc]\n",
    "  #Esta información representa lo procesado en cada capa.\n",
    "  out = [(None, _x)] #Es el caso inicial, ya que en la primera capa no hay iteración.\n",
    "  \n",
    "  #Forward pass, básicamente ir ejecutando capa por capa nuestra red, pasando el vector de entrada\n",
    "  for indiceL, L in enumerate(rn):\n",
    "    #Y recordar que el hacer esto, es aplicar una suma ponderada seguida de la función de activación\n",
    "    #Dicha suma ponderada, la llamaremos \"z\":\n",
    "    z = out[-1][1] @ rn[indiceL].w + rn[indiceL].b\n",
    "    #z = _x @ rn[0].w + rn[0].b #Así sería de manera individual con \"_x\".\n",
    "    #z = out[0][1] @ rn[0].w + rn[0].b #Y así con \"out\".\n",
    "    #Es un producto matricial: \"_x\" @ \"w\"\n",
    "    \n",
    "    #Activación:\n",
    "    a = rn[indiceL].act_f[0](z) #Recordar que nuestra act_f es una tupla con dos valores\n",
    "    #a = rn[0].act_f[0](z) #Así sería de manera individual.\n",
    "    #Todo lo guardamos en out:\n",
    "    out.append((z, a))\n",
    "\n",
    "  #print(out)\n",
    "  #print(\"---------------------------------------------\")\n",
    "  #print(out[-1][1]) #Con esto vemos el primera elemento del úlitmo resultado\n",
    "  #y dicho elemento es justamente la última matriz generada ya con todos los datos.\n",
    "  \n",
    "  #print(func_Coste[0](out[-1][1], _y)) #Con esto podemos visualizar cuanto es el error promedio.\n",
    "\n",
    "  if train: #Esto parte si es el entrenamiento como tal, lo anterior es solo para obtener un resultado.\n",
    "    #Backward pass\n",
    "    deltas = [] #Este será el vector con los errores calculados a partir de derivadas parciales.\n",
    "\n",
    "    for iL in reversed(range(0, len(rn))):\n",
    "      _z=out[iL+1][0]\n",
    "      _a=out[iL+1][1]\n",
    "      #print(_z.shape)\n",
    "      #print(_a.shape)\n",
    "\n",
    "      #Ya que para la última capa se le aplica una derivada un poco diferente a las demás\n",
    "      #Es que se hace la siguiente condición:\n",
    "      if iL == len(rn)-1:\n",
    "        #Calcular delta (derivada) en la última capa:\n",
    "        deltas.insert(0, func_Coste[1](_a, _y)*rn[iL].act_f[1](_a))\n",
    "        #_w = rn[iL].w\n",
    "      else:\n",
    "        #Calcular delta (derivada) respecto a capa previa:\n",
    "        deltas.insert(0, deltas[0]@_w.T * rn[iL].act_f[1](_a))\n",
    "        #Se inserta siempre en 0 para ir desplazando los demás resultados.\n",
    "        #deltas.insert(0, _w@deltas[0] * rn[iL].act_f[1](_a))\n",
    "\n",
    "      _w = rn[iL].w #Con esto vamos actualizando nuestros pesos.\n",
    "\n",
    "      #Gradient descent (Recordar que el gradiente es el vector que contiene a las pendientes)\n",
    "      #np.mean: calcula medias | axis: le dice que eje, del (500, 1) el 500\n",
    "      #keepdims: para que no lo convierta en serie, ejemplo de serie: (500, )\n",
    "      #y \"lr\" el cual es los pasos que daremos por cada iteración, ya lo vimos en el capítulo anterior.\n",
    "      rn[iL].b = rn[iL].b - np.mean(deltas[0], axis=0, keepdims=True) * lr #Básicamente: theta=theta-gradiente*lr\n",
    "      #Y de manera similar con \"w\":\n",
    "      rn[iL].w = rn[iL].w - out[iL][1].T@deltas[0] * lr #Con esto comenzamos a distribuir el error a las demás capas.\n",
    "\n",
    "  return out[-1][1] #Retornamos la última matriz calculada (u operada) que corresponde a la función de activación.\n",
    "\n",
    "#entrenar(red_neuronal, x, y, i2_cost, 0.5, train=True) #Línea que comprueba que está funcionando la función \"entrenar\"."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## A graficar!\n",
    "Ahora lo que haremos será instanciar una nueva red e ir graficando cada cierta cantidad de iteraciones, para así ir viendo como evoluciona nuestra red neuronal."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import time\n",
    "from IPython.display import clear_output\n",
    "\n",
    "#p definido al principio de este documento, al igual que x e y\n",
    "topologia = [p, 4, 1]\n",
    "#Función de activación definida desupués de la clase \"neural_layer\":\n",
    "redNeuronal = crear_redNeuronal(topologia, sigmoide)\n",
    "\n",
    "#Matriz que almacena cada variación entre la salida de la red y lo que esperamos.\n",
    "loss=[]\n",
    "\n",
    "for it in range(10000):\n",
    "  #Entrenamos a la red:\n",
    "  pY = entrenar(redNeuronal, x, y, i2_cost, lr=0.03)\n",
    "  #Con \"pY\" podremos comparar la salida de nuestra Red Neuronal con los valores que esperamos realmemente.\n",
    "\n",
    "  if it%25 == 0: #Cada 25 iteraciones o ciclos, vemos:\n",
    "    loss.append(i2_cost[0](pY, y))\n",
    "\n",
    "    #Resolución de nuestro gráfico:\n",
    "    res=50\n",
    "    #Valores aleatorios para hacer la gráfica base:\n",
    "    _x0 = np.linspace(-1.5, 1.5, res)\n",
    "    _x1 = np.linspace(-1.5, 1.5, res)\n",
    "    _Y = np.zeros((res, res))\n",
    "\n",
    "    #Ahora guardaremos\n",
    "    for ix0, x0 in enumerate(_x0):\n",
    "      for ix1, x1 in enumerate(_x1):\n",
    "        _Y[ix0, ix1] = entrenar(redNeuronal, np.array([[x0, x1]]), y, i2_cost, train=False)[0][0] #Aquí no entrenamos la red.\n",
    "        #Solo es para hacer predicciones de como la red neuronal está clasificando ambos conjuntos de puntos.\n",
    "\n",
    "    #Aquí comenzamos creando la gráfica, como en el anterior ejercicio:\n",
    "    plt.pcolormesh(_x0, _x1, _Y, cmap=\"coolwarm\") #Coloreando superficies.\n",
    "    plt.axis(\"equal\") #Proporcionamos los ejes.\n",
    "\n",
    "    #Esto está al principio y es para visualizar nuestro conjunto de datos:\n",
    "    plt.scatter(x[y[:, 0]==0, 0], x[y[:, 0]==0, 1], c=\"skyblue\")\n",
    "    plt.scatter(x[y[:, 0]==1, 0], x[y[:, 0]==1, 1], c=\"salmon\")\n",
    "\n",
    "    #No explico mucho pero básicamente borra y vuelve a hacer el dibujo dandole un efecto como de animado:\n",
    "    clear_output(wait=True)\n",
    "    plt.show() #Mostramos el primer gráfico con las superficies y nuestros puntos.\n",
    "    plt.plot(range(len(loss)), loss)\n",
    "    plt.show() #Mostramos un segundo gráfico de como el error se va comportando.\n",
    "    time.sleep(0.1) #Y un pequeño tiempo de descanso para la CPU xD"
   ]
  }
 ],
 "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}