Example/helloWorld.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hello World Example\n",
    "\n",
    "This notebook guides through the generation of the example\n",
    "`NeuralNetwork.Examples.HelloWorld` and will generate a neural network to\n",
    "replace `NeuralNetwork.Examples.Utilities.SimpleEquation`\n",
    "\n",
    "```modelica\n",
    "block SimpleEquation\n",
    "  extends Modelica.Blocks.Interfaces.SISO;\n",
    "equation\n",
    "  y = u*u + 0.5*u - 2.0;\n",
    "end SimpleEquation;\n",
    "```\n",
    "\n",
    "with a single input single output dense feed forward neural network."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's start by generating a training data set for\n",
    "\n",
    "$$ f: [-1, 1] \\to \\mathbb{R}, f(u) = u^2 + 0.5 u - 2  $$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "\n",
    "# Deterministic results\n",
    "tf.keras.utils.set_random_seed(42)  # sets seeds for base-python, numpy and tf\n",
    "tf.config.experimental.enable_op_determinism()\n",
    "\n",
    "def f(u):\n",
    "  return u**2 + 0.5*u - 2.0\n",
    "\n",
    "# Data\n",
    "x = np.linspace(start=-1.0, stop=1.0, num=10000)\n",
    "y = f(x)\n",
    "\n",
    "data_train, data_test = tf.keras.utils.split_dataset(\n",
    "    (x,y), left_size=0.8, right_size=0.2, shuffle=True, seed=None\n",
    ")\n",
    "\n",
    "batch_size = 32\n",
    "data_train_batches = data_train.batch(batch_size)\n",
    "data_test_batches = data_test.batch(batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we define a sequential neural network `model` with one input layer (two\n",
    "neurons) and one output layer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\">Model: \"sequential\"</span>\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1mModel: \"sequential\"\u001b[0m\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃<span style=\"font-weight: bold\"> Layer (type)                    </span>┃<span style=\"font-weight: bold\"> Output Shape           </span>┃<span style=\"font-weight: bold\">       Param # </span>┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ dense (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                   │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">2</span>)              │             <span style=\"color: #00af00; text-decoration-color: #00af00\">4</span> │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_1 (<span style=\"color: #0087ff; text-decoration-color: #0087ff\">Dense</span>)                 │ (<span style=\"color: #00d7ff; text-decoration-color: #00d7ff\">None</span>, <span style=\"color: #00af00; text-decoration-color: #00af00\">1</span>)              │             <span style=\"color: #00af00; text-decoration-color: #00af00\">3</span> │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n",
       "</pre>\n"
      ],
      "text/plain": [
       "┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓\n",
       "┃\u001b[1m \u001b[0m\u001b[1mLayer (type)                   \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1mOutput Shape          \u001b[0m\u001b[1m \u001b[0m┃\u001b[1m \u001b[0m\u001b[1m      Param #\u001b[0m\u001b[1m \u001b[0m┃\n",
       "┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩\n",
       "│ dense (\u001b[38;5;33mDense\u001b[0m)                   │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m2\u001b[0m)              │             \u001b[38;5;34m4\u001b[0m │\n",
       "├─────────────────────────────────┼────────────────────────┼───────────────┤\n",
       "│ dense_1 (\u001b[38;5;33mDense\u001b[0m)                 │ (\u001b[38;5;45mNone\u001b[0m, \u001b[38;5;34m1\u001b[0m)              │             \u001b[38;5;34m3\u001b[0m │\n",
       "└─────────────────────────────────┴────────────────────────┴───────────────┘\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Total params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">7</span> (28.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Total params: \u001b[0m\u001b[38;5;34m7\u001b[0m (28.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">7</span> (28.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Trainable params: \u001b[0m\u001b[38;5;34m7\u001b[0m (28.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"font-weight: bold\"> Non-trainable params: </span><span style=\"color: #00af00; text-decoration-color: #00af00\">0</span> (0.00 B)\n",
       "</pre>\n"
      ],
      "text/plain": [
       "\u001b[1m Non-trainable params: \u001b[0m\u001b[38;5;34m0\u001b[0m (0.00 B)\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential([\n",
    "  tf.keras.Input(shape=(1,)),\n",
    "  tf.keras.layers.Dense(2, activation='tanh'),\n",
    "  tf.keras.layers.Dense(1, activation='linear')\n",
    "])\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we compile the model and train it on the training data set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer = 'adam',\n",
    "              loss = tf.keras.losses.MeanSquaredError,\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 589us/step - accuracy: 0.0000e+00 - loss: 2.9083\n",
      "Epoch 2/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 474us/step - accuracy: 0.0000e+00 - loss: 0.8081\n",
      "Epoch 3/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 453us/step - accuracy: 0.0000e+00 - loss: 0.1659\n",
      "Epoch 4/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 489us/step - accuracy: 0.0000e+00 - loss: 0.0703\n",
      "Epoch 5/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 467us/step - accuracy: 0.0000e+00 - loss: 0.0585\n",
      "Epoch 6/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 474us/step - accuracy: 0.0000e+00 - loss: 0.0545\n",
      "Epoch 7/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 485us/step - accuracy: 0.0000e+00 - loss: 0.0501\n",
      "Epoch 8/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 538us/step - accuracy: 0.0000e+00 - loss: 0.0447\n",
      "Epoch 9/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 478us/step - accuracy: 0.0000e+00 - loss: 0.0384\n",
      "Epoch 10/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 458us/step - accuracy: 0.0000e+00 - loss: 0.0315\n",
      "Epoch 11/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 472us/step - accuracy: 0.0000e+00 - loss: 0.0244\n",
      "Epoch 12/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 476us/step - accuracy: 0.0000e+00 - loss: 0.0176\n",
      "Epoch 13/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 472us/step - accuracy: 0.0000e+00 - loss: 0.0119\n",
      "Epoch 14/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 483us/step - accuracy: 0.0000e+00 - loss: 0.0075\n",
      "Epoch 15/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 480us/step - accuracy: 0.0000e+00 - loss: 0.0046\n",
      "Epoch 16/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 540us/step - accuracy: 0.0000e+00 - loss: 0.0028\n",
      "Epoch 17/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 506us/step - accuracy: 0.0000e+00 - loss: 0.0020\n",
      "Epoch 18/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 490us/step - accuracy: 0.0000e+00 - loss: 0.0016\n",
      "Epoch 19/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 469us/step - accuracy: 0.0000e+00 - loss: 0.0014\n",
      "Epoch 20/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 474us/step - accuracy: 0.0000e+00 - loss: 0.0013\n",
      "Epoch 21/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 528us/step - accuracy: 0.0000e+00 - loss: 0.0013\n",
      "Epoch 22/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 651us/step - accuracy: 0.0000e+00 - loss: 0.0012\n",
      "Epoch 23/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 499us/step - accuracy: 0.0000e+00 - loss: 0.0012\n",
      "Epoch 24/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 494us/step - accuracy: 0.0000e+00 - loss: 0.0012\n",
      "Epoch 25/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 482us/step - accuracy: 0.0000e+00 - loss: 0.0012\n",
      "Epoch 26/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 501us/step - accuracy: 0.0000e+00 - loss: 0.0011\n",
      "Epoch 27/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 479us/step - accuracy: 0.0000e+00 - loss: 0.0011\n",
      "Epoch 28/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 487us/step - accuracy: 0.0000e+00 - loss: 0.0011\n",
      "Epoch 29/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 477us/step - accuracy: 0.0000e+00 - loss: 0.0011\n",
      "Epoch 30/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 480us/step - accuracy: 0.0000e+00 - loss: 0.0011\n",
      "Epoch 31/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 471us/step - accuracy: 0.0000e+00 - loss: 0.0010\n",
      "Epoch 32/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 487us/step - accuracy: 0.0000e+00 - loss: 0.0010\n",
      "Epoch 33/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 489us/step - accuracy: 0.0000e+00 - loss: 0.0010\n",
      "Epoch 34/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 483us/step - accuracy: 0.0000e+00 - loss: 9.8657e-04\n",
      "Epoch 35/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 475us/step - accuracy: 0.0000e+00 - loss: 9.6779e-04\n",
      "Epoch 36/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 471us/step - accuracy: 0.0000e+00 - loss: 9.4862e-04\n",
      "Epoch 37/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 484us/step - accuracy: 0.0000e+00 - loss: 9.2910e-04\n",
      "Epoch 38/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 472us/step - accuracy: 0.0000e+00 - loss: 9.0933e-04\n",
      "Epoch 39/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 484us/step - accuracy: 0.0000e+00 - loss: 8.8939e-04\n",
      "Epoch 40/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 472us/step - accuracy: 0.0000e+00 - loss: 8.6939e-04\n",
      "Epoch 41/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 470us/step - accuracy: 0.0000e+00 - loss: 8.4937e-04\n",
      "Epoch 42/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 468us/step - accuracy: 0.0000e+00 - loss: 8.2943e-04\n",
      "Epoch 43/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 467us/step - accuracy: 0.0000e+00 - loss: 8.0965e-04\n",
      "Epoch 44/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 473us/step - accuracy: 0.0000e+00 - loss: 7.9012e-04\n",
      "Epoch 45/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 491us/step - accuracy: 0.0000e+00 - loss: 7.7093e-04\n",
      "Epoch 46/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 479us/step - accuracy: 0.0000e+00 - loss: 7.5217e-04\n",
      "Epoch 47/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 458us/step - accuracy: 0.0000e+00 - loss: 7.3389e-04\n",
      "Epoch 48/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 467us/step - accuracy: 0.0000e+00 - loss: 7.1612e-04\n",
      "Epoch 49/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 470us/step - accuracy: 0.0000e+00 - loss: 6.9889e-04\n",
      "Epoch 50/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 472us/step - accuracy: 0.0000e+00 - loss: 6.8220e-04\n",
      "Epoch 51/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 495us/step - accuracy: 0.0000e+00 - loss: 6.6605e-04\n",
      "Epoch 52/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 455us/step - accuracy: 0.0000e+00 - loss: 6.5043e-04\n",
      "Epoch 53/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 455us/step - accuracy: 0.0000e+00 - loss: 6.3532e-04\n",
      "Epoch 54/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 473us/step - accuracy: 0.0000e+00 - loss: 6.2069e-04\n",
      "Epoch 55/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 482us/step - accuracy: 0.0000e+00 - loss: 6.0653e-04\n",
      "Epoch 56/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 467us/step - accuracy: 0.0000e+00 - loss: 5.9281e-04\n",
      "Epoch 57/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 535us/step - accuracy: 0.0000e+00 - loss: 5.7951e-04\n",
      "Epoch 58/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 480us/step - accuracy: 0.0000e+00 - loss: 5.6660e-04\n",
      "Epoch 59/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 483us/step - accuracy: 0.0000e+00 - loss: 5.5407e-04\n",
      "Epoch 60/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 479us/step - accuracy: 0.0000e+00 - loss: 5.4189e-04\n",
      "Epoch 61/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 473us/step - accuracy: 0.0000e+00 - loss: 5.3004e-04\n",
      "Epoch 62/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 489us/step - accuracy: 0.0000e+00 - loss: 5.1850e-04\n",
      "Epoch 63/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 482us/step - accuracy: 0.0000e+00 - loss: 5.0727e-04\n",
      "Epoch 64/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 471us/step - accuracy: 0.0000e+00 - loss: 4.9631e-04\n",
      "Epoch 65/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 453us/step - accuracy: 0.0000e+00 - loss: 4.8562e-04\n",
      "Epoch 66/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 461us/step - accuracy: 0.0000e+00 - loss: 4.7518e-04\n",
      "Epoch 67/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 616us/step - accuracy: 0.0000e+00 - loss: 4.6498e-04\n",
      "Epoch 68/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 541us/step - accuracy: 0.0000e+00 - loss: 4.5501e-04\n",
      "Epoch 69/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 484us/step - accuracy: 0.0000e+00 - loss: 4.4525e-04\n",
      "Epoch 70/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 495us/step - accuracy: 0.0000e+00 - loss: 4.3570e-04\n",
      "Epoch 71/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 490us/step - accuracy: 0.0000e+00 - loss: 4.2635e-04\n",
      "Epoch 72/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 469us/step - accuracy: 0.0000e+00 - loss: 4.1718e-04\n",
      "Epoch 73/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 467us/step - accuracy: 0.0000e+00 - loss: 4.0818e-04\n",
      "Epoch 74/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 491us/step - accuracy: 0.0000e+00 - loss: 3.9936e-04\n",
      "Epoch 75/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 473us/step - accuracy: 0.0000e+00 - loss: 3.9070e-04\n",
      "Epoch 76/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 453us/step - accuracy: 0.0000e+00 - loss: 3.8219e-04\n",
      "Epoch 77/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 453us/step - accuracy: 0.0000e+00 - loss: 3.7383e-04\n",
      "Epoch 78/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 463us/step - accuracy: 0.0000e+00 - loss: 3.6561e-04\n",
      "Epoch 79/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 465us/step - accuracy: 0.0000e+00 - loss: 3.5753e-04\n",
      "Epoch 80/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 447us/step - accuracy: 0.0000e+00 - loss: 3.4958e-04\n",
      "Epoch 81/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 457us/step - accuracy: 0.0000e+00 - loss: 3.4175e-04\n",
      "Epoch 82/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 446us/step - accuracy: 0.0000e+00 - loss: 3.3404e-04\n",
      "Epoch 83/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 459us/step - accuracy: 0.0000e+00 - loss: 3.2645e-04\n",
      "Epoch 84/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 483us/step - accuracy: 0.0000e+00 - loss: 3.1896e-04\n",
      "Epoch 85/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 484us/step - accuracy: 0.0000e+00 - loss: 3.1158e-04\n",
      "Epoch 86/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 469us/step - accuracy: 0.0000e+00 - loss: 3.0431e-04\n",
      "Epoch 87/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 476us/step - accuracy: 0.0000e+00 - loss: 2.9713e-04\n",
      "Epoch 88/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 507us/step - accuracy: 0.0000e+00 - loss: 2.9004e-04\n",
      "Epoch 89/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 577us/step - accuracy: 0.0000e+00 - loss: 2.8305e-04\n",
      "Epoch 90/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 513us/step - accuracy: 0.0000e+00 - loss: 2.7614e-04\n",
      "Epoch 91/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 476us/step - accuracy: 0.0000e+00 - loss: 2.6932e-04\n",
      "Epoch 92/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 462us/step - accuracy: 0.0000e+00 - loss: 2.6257e-04\n",
      "Epoch 93/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 470us/step - accuracy: 0.0000e+00 - loss: 2.5591e-04\n",
      "Epoch 94/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 456us/step - accuracy: 0.0000e+00 - loss: 2.4933e-04\n",
      "Epoch 95/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 491us/step - accuracy: 0.0000e+00 - loss: 2.4281e-04\n",
      "Epoch 96/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 478us/step - accuracy: 0.0000e+00 - loss: 2.3638e-04\n",
      "Epoch 97/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 612us/step - accuracy: 0.0000e+00 - loss: 2.3001e-04\n",
      "Epoch 98/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 468us/step - accuracy: 0.0000e+00 - loss: 2.2371e-04\n",
      "Epoch 99/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 478us/step - accuracy: 0.0000e+00 - loss: 2.1748e-04\n",
      "Epoch 100/100\n",
      "\u001b[1m250/250\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 507us/step - accuracy: 0.0000e+00 - loss: 2.1132e-04\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.history.History at 0x239fb1357f0>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(data_train_batches, epochs=100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can plot the prediction of the model to verify that the neural network is\n",
    "sufficiently trained."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "ax.plot(x, model(x), c='tab:blue', label='prediction', alpha=0.3)\n",
    "ax.plot(x, y, c='tab:orange', label='reference', alpha=0.3)\n",
    "ax.legend()\n",
    "ax.grid(True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Extract weights and biases\n",
    "\n",
    "To export the neural network from TensorFlow to Modelica we need to extract\n",
    "weights and biases for each layer and generate corresponding Modelica code.\n",
    "\n",
    "Remember that TensorFlow uses weight representation\n",
    "\n",
    "$$ y = f(x W^T + b) $$\n",
    "\n",
    "to be fast on large data sets, but NeuralNetwork library expects format\n",
    "\n",
    "$$ y = f(W x + b). $$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights_and_biases = []\n",
    "for layer in model.layers:\n",
    "    layer_weights = layer.get_weights()\n",
    "    weights_and_biases.append(layer_weights)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Install Python module pythontomodelica locally by running\n",
    "\n",
    "```bash\n",
    "pip install -e ./PythonToModelica\n",
    "```\n",
    "\n",
    "With `to_modelica` we can convert NumPy N-dimensional arrays to Modelica arrays."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "parameter Real[2,1] layer_1_weights = {{-0.95248}, {-0.943175}};\n",
      "parameter Real[2] layer_1_bias = {0.872633, -0.949252};\n",
      "parameter Real[1,2] layer_2_weights = {{-2.25385, 1.40389}};\n",
      "parameter Real[1] layer_2_bias = {0.60548};\n"
     ]
    }
   ],
   "source": [
    "from pythontomodelica import to_modelica\n",
    "\n",
    "modelica_string = []\n",
    "for i, layer_weights in enumerate(weights_and_biases):\n",
    "  weights, bias = layer_weights\n",
    "  modelica_string.append(\"parameter \" + to_modelica(f\"layer_{i+1}_weights\", weights.transpose()))\n",
    "  modelica_string.append(\"parameter \" + to_modelica(f\"layer_{i+1}_bias\", bias))\n",
    "\n",
    "print(\"\\n\".join(modelica_string))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally we need to copy these parameters into our new Modelica model\n",
    "`NeuralNetwork.Examples.Utilities.SimpleNetwork` and set the number of inputs,\n",
    "outputs and neurons for the network and each layer.\n",
    "Finally we connect:\n",
    "  - the input to the input layer \n",
    "  - the input layer to the output layer\n",
    "  - the output layer to the output"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```modelica\n",
    "block SimpleNetwork \"Neural Network approximating y = u*u + 0.5*u - 2.0 on interval [-1,1]\"\n",
    "  extends NeuralNetwork.Networks.Interfaces.Network(numInputs = 1, numOutputs = 1);\n",
    "  Layer.Input inputLayer(\n",
    "    numInputs = 1,\n",
    "    numNeurons = 2,\n",
    "    weights = layer_1_weights,\n",
    "    bias = layer_1_bias,\n",
    "    redeclare function f = NeuralNetwork.ActivationFunctions.Tanh\n",
    "  );\n",
    "  Layer.Output outputLayer(\n",
    "    numInputs = 2,\n",
    "    numNeurons = 1,\n",
    "    weights = layer_2_weights,\n",
    "    bias = layer_2_bias\n",
    "  );\n",
    "\n",
    "  parameter Real[2,1] layer_1_weights = {{-0.95248}, {-0.943175}};\n",
    "  parameter Real[2] layer_1_bias = {0.872633, -0.949252};\n",
    "  parameter Real[1,2] layer_2_weights = {{-2.25385, 1.40389}};\n",
    "  parameter Real[1] layer_2_bias = {0.60548};\n",
    "equation\n",
    "  connect(u, inputLayer.u);\n",
    "  connect(inputLayer.y, outputLayer.u);\n",
    "  connect(outputLayer.y, y);\n",
    "end SimpleNetwork;\n",
    "```"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.12.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}