{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.python.framework import ops"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "def lrelu(x, alpha=0.01):\n",
    "    return np.maximum(alpha*x,x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "a=6.12"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {},
   "outputs": [],
   "source": [
    "def HCAL_act(x):\n",
    "    \n",
    "\n",
    "    return lrelu(np.minimum(2.+0.01*x,x))\n",
    "    \n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0"
      ]
     },
     "execution_count": 197,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "HCAL_act(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {},
   "outputs": [],
   "source": [
    "np_HCAL_act = np.vectorize(HCAL_act)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<BarContainer object of 200 artists>"
      ]
     },
     "execution_count": 199,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.bar(np.arange(-10,10,0.1),np_HCAL_act(np.arange(-10,10,0.1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def d_HCAL_act(x):\n",
    "    num = np.multiply(a,np.exp(np.add(a,-x)))\n",
    "    den = np.square(np.add(1, np.exp(np.add(a,-x))))\n",
    "    return num/den"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "np_d_HCAL_act = np.vectorize(d_HCAL_act)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<BarContainer object of 300 artists>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.bar(np.arange(-10,20,0.1),np_d_HCAL_act(np.arange(-10,20,0.1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "np_d_HCAL_act_32 = lambda x: np_d_HCAL_act(x).astype(np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def tf_d_HCAL_act(x,name=None):\n",
    "    with tf.name_scope(name, \"d_HCAL_act\", [x]) as name:\n",
    "        y = tf.py_func(np_d_HCAL_act_32,\n",
    "                        [x],\n",
    "                        [tf.float32],\n",
    "                        name=name,\n",
    "                        stateful=False)\n",
    "        return y[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def py_func(func, inp, Tout, stateful=True, name=None, grad=None):\n",
    "\n",
    "    # Need to generate a unique name to avoid duplicates:\n",
    "    rnd_name = 'PyFuncGrad' + str(np.random.randint(0, 1E+8))\n",
    "\n",
    "    tf.RegisterGradient(rnd_name)(grad)  # see _MySquareGrad for grad example\n",
    "    g = tf.get_default_graph()\n",
    "    with g.gradient_override_map({\"PyFunc\": rnd_name}):\n",
    "        return tf.py_func(func, inp, Tout, stateful=stateful, name=name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def HCAL_grad(op, grad):\n",
    "    x = op.inputs[0]\n",
    "\n",
    "    n_gr = tf_d_HCAL_act(x)\n",
    "    return grad * n_gr "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "np_HCAL_act_32 = lambda x: np_HCAL_act(x).astype(np.float32)\n",
    "\n",
    "def tf_HCAL_act(x, name=None):\n",
    "\n",
    "    with tf.name_scope(name, \"HCAL_act\", [x]) as name:\n",
    "        y = py_func(np_HCAL_act_32,\n",
    "                        [x],\n",
    "                        [tf.float32],\n",
    "                        name=name,\n",
    "                        grad=HCAL_grad)  # <-- here's the call to the gradient\n",
    "        return y[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def tf_HCAL_act(x):\n",
    "    a=6.12\n",
    "    arg=tf.cast(-x,tf.float32)\n",
    "    return tf.divide(a,tf.add(1.,tf.exp(tf.add(a,arg))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/hep/davide/miniconda3/envs/deepnet/lib/python3.6/site-packages/tensorflow/python/util/tf_should_use.py:107: initialize_all_variables (from tensorflow.python.ops.variables) is deprecated and will be removed after 2017-03-02.\n",
      "Instructions for updating:\n",
      "Use `tf.global_variables_initializer` instead.\n",
      "[-10.  -9.  -8.  -7.  -6.  -5.  -4.  -3.  -2.  -1.   0.   1.   2.   3.\n",
      "   4.   5.   6.   7.   8.   9.] [6.1083620e-07 1.6604231e-06 4.5134962e-06 1.2268938e-05 3.3350316e-05\n",
      " 9.0654714e-05 2.4641879e-04 6.6978938e-04 1.8203339e-03 4.9456530e-03\n",
      " 1.3425037e-02 3.6355998e-02 9.7827286e-02 2.5881341e-01 6.5586865e-01\n",
      " 1.5055892e+00 2.8766201e+00 4.3257518e+00 5.3097801e+00 5.7947154e+00] [6.10836139e-07 1.66042264e-06 4.51349297e-06 1.22689135e-05\n",
      " 3.33501339e-05 9.06533678e-05 2.46408861e-04 6.69716042e-04\n",
      " 1.81979244e-03 4.94165625e-03 1.33955870e-02 3.61400247e-02\n",
      " 9.62635279e-02 2.47868240e-01 5.85580468e-01 1.13519740e+00\n",
      " 1.52450514e+00 1.26821423e+00 7.02955782e-01 3.07995707e-01]\n"
     ]
    }
   ],
   "source": [
    "with tf.Session() as sess:\n",
    "\n",
    "    x = tf.constant(np.arange(-10,10,1.))\n",
    "    y = tf_HCAL_act(x)\n",
    "    tf.initialize_all_variables().run()\n",
    "\n",
    "    print(x.eval(), y.eval(), tf.gradients(y, [x])[0].eval())\n"
   ]
  },
  {
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}