Variational_Classifier_Iris.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyMM7rB2oybg8IL/ExtxqWts",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/MonitSharma/Learn-Quantum-Machine-Learning/blob/main/Variational_Classifier_Iris.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Variational Classifier\n",
        "\n",
        "Here, we will learn how to implement variational quantum classifiers, that can be trained from labelled data to classify new data samples.\n",
        "\n",
        "Work inspired from [Farhi and Neven (2018)](https://arxiv.org/abs/1802.06002) and [Schuld et al. (2018)](https://arxiv.org/abs/1804.00633)\n",
        "\n",
        "This tutorial is taken from the Pennylanes handbook."
      ],
      "metadata": {
        "id": "nyBdpM_PTqFm"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "How a variational quantum classifier reproduce the parity function\n",
        "\n",
        "$$ \\begin{split}f: x \\in \\{0,1\\}^{\\otimes n} \\rightarrow y =\n",
        "\\begin{cases} 1 \\text{  if uneven number of ones in } x \\\\ 0\n",
        "\\text{ otherwise} \\end{cases}.\\end{split}$$\n",
        "\n",
        "This examples, demonstrates how to encode binary inputs into the initial state of the variational circuit, which is simply the computational basis state.\n",
        "\n"
      ],
      "metadata": {
        "id": "m1-L0TWAUGPC"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Fitting the parity function\n",
        "\n",
        "\n",
        "### Imports\n",
        "\n",
        "Import PennyLane and their version of NumPy and an optimizer"
      ],
      "metadata": {
        "id": "rfdCZREoUZP8"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "Q7tXI_3MTmjW"
      },
      "outputs": [],
      "source": [
        "import pennylane as qml\n",
        "from pennylane import numpy as np\n",
        "from pennylane.optimize import NesterovMomentumOptimizer\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Quantum and Classical Nodes\n",
        "\n",
        "We create a quantum device with four qubits"
      ],
      "metadata": {
        "id": "pjhjmMmyUutH"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "dev = qml.device(\"default.qubit\", wires=4)"
      ],
      "metadata": {
        "id": "Ub_NJm6TUtAq"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Variational classifiers usually define a layer, which is elementary circuit architecture that gets repeated to build the variational circuit.\n",
        "\n",
        "\n",
        "Our circuit layers consists of an arbitrary rotation on every qubit, as well as CNOTs that entangle each qubit with its neighbour"
      ],
      "metadata": {
        "id": "rcr94YQWU57n"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def layer(W):\n",
        "\n",
        "    qml.Rot(W[0, 0], W[0, 1], W[0, 2], wires=0)\n",
        "    qml.Rot(W[1, 0], W[1, 1], W[1, 2], wires=1)\n",
        "    qml.Rot(W[2, 0], W[2, 1], W[2, 2], wires=2)\n",
        "    qml.Rot(W[3, 0], W[3, 1], W[3, 2], wires=3)\n",
        "\n",
        "    qml.CNOT(wires=[0, 1])\n",
        "    qml.CNOT(wires=[1, 2])\n",
        "    qml.CNOT(wires=[2, 3])\n",
        "    qml.CNOT(wires=[3, 0])"
      ],
      "metadata": {
        "id": "lwsrMJsGU5XA"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "A Rot gate is used here, which is\n",
        "\n",
        "![rottt (1).png](data:image/png;base64,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)"
      ],
      "metadata": {
        "id": "SFQKuMStVQMr"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "We also need a way to encode data inputs \n",
        "$x$\n",
        " into the circuit, so that the measured output depends on the inputs. In this first example, the inputs are bitstrings, which we encode into the state of the qubits. The quantum state \n",
        "$ψ$\n",
        " after state preparation is a computational basis state that has 1s where \n",
        "$x$\n",
        " has 1s, for example\n",
        "\n",
        "$$x = 0101 \\rightarrow |\\psi \\rangle = |0101 \\rangle .$$\n",
        "\n",
        "We use the `BasisState` function rovided by PennyLane, which expects `x` to be a list of zeros and ones, i.e `[0,1,0,1]`\n"
      ],
      "metadata": {
        "id": "j3thSmkcWGlG"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def statepreparation(x):\n",
        "    qml.BasisState(x, wires=[0, 1, 2, 3])"
      ],
      "metadata": {
        "id": "jah7nY4lVOdT"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Now we define the quantum node as a state preparation routine, followed by a repetition of the layer structure. Borrowing from machine learning, we call the parameters `weights`."
      ],
      "metadata": {
        "id": "VR31BoAFWigI"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "@qml.qnode(dev)\n",
        "def circuit(weights, x):\n",
        "\n",
        "    statepreparation(x)\n",
        "\n",
        "    for W in weights:\n",
        "        layer(W)\n",
        "\n",
        "    return qml.expval(qml.PauliZ(0))"
      ],
      "metadata": {
        "id": "l7Lo9KOTWhzu"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Different from previous examples, the quantum node takes the data as a keyword argument `x` (with the default value `None`). Keyword arguments of a quantum node are considered as fixed when calculating a gradient; they are never trained.\n",
        "\n",
        "If we want to add a “classical” bias parameter, the variational quantum classifier also needs some post-processing. We define the final model by a classical node that uses the first variable, and feeds the remainder into the quantum node. Before this, we reshape the list of remaining variables for easy use in the quantum node."
      ],
      "metadata": {
        "id": "ridbzC25Wqz3"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def variational_classifier(weights, bias, x):\n",
        "    return circuit(weights, x) + bias\n"
      ],
      "metadata": {
        "id": "FpsBUTP5WnwB"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Cost\n",
        "In supervised learning, the cost function is usually the sum of a loss function and a regularizer. We use the standard square loss that measures the distance between target labels and model predictions."
      ],
      "metadata": {
        "id": "7zc2x6noWw_R"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def square_loss(labels, predictions):\n",
        "    loss = 0\n",
        "    for l, p in zip(labels, predictions):\n",
        "        loss = loss + (l - p) ** 2\n",
        "\n",
        "    loss = loss / len(labels)\n",
        "    return loss"
      ],
      "metadata": {
        "id": "GauW2gH5WwT1"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "To monitor how many inputs the current classifier predicted correctly, we also define the accuracy given target labels and model predictions."
      ],
      "metadata": {
        "id": "QW0wrT3-W3sK"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def accuracy(labels, predictions):\n",
        "\n",
        "    loss = 0\n",
        "    for l, p in zip(labels, predictions):\n",
        "        if abs(l - p) < 1e-5:\n",
        "            loss = loss + 1\n",
        "    loss = loss / len(labels)\n",
        "\n",
        "    return loss"
      ],
      "metadata": {
        "id": "pBIngpm6W16m"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "For learning tasks, the cost depends on the data - here the features and labels considered in the iteration of the optimization routine."
      ],
      "metadata": {
        "id": "Yuky7P6OW7jU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def cost(weights, bias, X, Y):\n",
        "    predictions = [variational_classifier(weights, bias, x) for x in X]\n",
        "    return square_loss(Y, predictions)\n"
      ],
      "metadata": {
        "id": "YRmWTfKiW58L"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Optimization\n",
        "\n",
        "Let's load and preprocess the data.\n",
        "\n",
        "Download the data from [here](https://raw.githubusercontent.com/XanaduAI/qml/master/demonstrations/variational_classifier/data/parity.txt)\n",
        "and place it in the subfolded `variational/data`\n",
        "\n"
      ],
      "metadata": {
        "id": "bVSwjH28XAcl"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "data = np.loadtxt(\"variational/data/parity.txt\")\n",
        "X = np.array(data[:, :-1], requires_grad=False)\n",
        "Y = np.array(data[:, -1], requires_grad=False)\n",
        "Y = Y * 2 - np.ones(len(Y))  # shift label from {0, 1} to {-1, 1}\n",
        "\n",
        "for i in range(5):\n",
        "    print(\"X = {}, Y = {: d}\".format(X[i], int(Y[i])))\n",
        "\n",
        "print(\"...\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JMpctTWTW93j",
        "outputId": "5d875f97-5ca5-4a09-e7f1-9e8d359b483f"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "X = [0. 0. 0. 0.], Y = -1\n",
            "X = [0. 0. 0. 1.], Y =  1\n",
            "X = [0. 0. 1. 0.], Y =  1\n",
            "X = [0. 0. 1. 1.], Y = -1\n",
            "X = [0. 1. 0. 0.], Y =  1\n",
            "...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "We initialize the variables randomly (but fix a seed for reproducibility). The first variable in the list is used as a bias, while the rest is fed into the gates of the variational circuit."
      ],
      "metadata": {
        "id": "THOcNME9X61X"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "np.random.seed(0)\n",
        "num_qubits = 4\n",
        "num_layers = 2\n",
        "weights_init = 0.01 * np.random.randn(num_layers, num_qubits, 3, requires_grad=True)\n",
        "bias_init = np.array(0.0, requires_grad=True)\n",
        "\n",
        "print(weights_init, bias_init)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bWlnpGmUXyGa",
        "outputId": "2aeed83f-8ae9-433b-811a-9a0c58cffd88"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[[[ 0.01764052  0.00400157  0.00978738]\n",
            "  [ 0.02240893  0.01867558 -0.00977278]\n",
            "  [ 0.00950088 -0.00151357 -0.00103219]\n",
            "  [ 0.00410599  0.00144044  0.01454274]]\n",
            "\n",
            " [[ 0.00761038  0.00121675  0.00443863]\n",
            "  [ 0.00333674  0.01494079 -0.00205158]\n",
            "  [ 0.00313068 -0.00854096 -0.0255299 ]\n",
            "  [ 0.00653619  0.00864436 -0.00742165]]] 0.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Next we create an optimizer and choose a batch size…"
      ],
      "metadata": {
        "id": "3Sp45MpNYBre"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "opt = NesterovMomentumOptimizer(0.5)\n",
        "batch_size = 5"
      ],
      "metadata": {
        "id": "U2nzqj46X-j1"
      },
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "…and train the optimizer. We track the accuracy - the share of correctly classified data samples. For this we compute the outputs of the variational classifier and turn them into predictions in $\\{−1,1\\}$\n",
        " by taking the sign of the output."
      ],
      "metadata": {
        "id": "FAh_XpU2YHTO"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "weights = weights_init\n",
        "bias = bias_init\n",
        "for it in range(25):\n",
        "\n",
        "    # Update the weights by one optimizer step\n",
        "    batch_index = np.random.randint(0, len(X), (batch_size,))\n",
        "    X_batch = X[batch_index]\n",
        "    Y_batch = Y[batch_index]\n",
        "    weights, bias, _, _ = opt.step(cost, weights, bias, X_batch, Y_batch)\n",
        "\n",
        "    # Compute accuracy\n",
        "    predictions = [np.sign(variational_classifier(weights, bias, x)) for x in X]\n",
        "    acc = accuracy(Y, predictions)\n",
        "\n",
        "    print(\n",
        "        \"Iter: {:5d} | Cost: {:0.7f} | Accuracy: {:0.7f} \".format(\n",
        "            it + 1, cost(weights, bias, X, Y), acc\n",
        "        )\n",
        "    )"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "aMX8JBv5YE6S",
        "outputId": "1c740dca-a4a1-4e40-d0a4-f3d00da0c13c"
      },
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Iter:     1 | Cost: 3.4355534 | Accuracy: 0.5000000 \n",
            "Iter:     2 | Cost: 1.9717733 | Accuracy: 0.5000000 \n",
            "Iter:     3 | Cost: 1.8182812 | Accuracy: 0.5000000 \n",
            "Iter:     4 | Cost: 1.5042404 | Accuracy: 0.5000000 \n",
            "Iter:     5 | Cost: 1.1477739 | Accuracy: 0.5000000 \n",
            "Iter:     6 | Cost: 1.2734990 | Accuracy: 0.6250000 \n",
            "Iter:     7 | Cost: 0.8290628 | Accuracy: 0.5000000 \n",
            "Iter:     8 | Cost: 0.3226183 | Accuracy: 1.0000000 \n",
            "Iter:     9 | Cost: 0.1436206 | Accuracy: 1.0000000 \n",
            "Iter:    10 | Cost: 0.2982810 | Accuracy: 1.0000000 \n",
            "Iter:    11 | Cost: 0.3064355 | Accuracy: 1.0000000 \n",
            "Iter:    12 | Cost: 0.1682335 | Accuracy: 1.0000000 \n",
            "Iter:    13 | Cost: 0.0892512 | Accuracy: 1.0000000 \n",
            "Iter:    14 | Cost: 0.0381562 | Accuracy: 1.0000000 \n",
            "Iter:    15 | Cost: 0.0170359 | Accuracy: 1.0000000 \n",
            "Iter:    16 | Cost: 0.0109353 | Accuracy: 1.0000000 \n",
            "Iter:    17 | Cost: 0.0108388 | Accuracy: 1.0000000 \n",
            "Iter:    18 | Cost: 0.0139196 | Accuracy: 1.0000000 \n",
            "Iter:    19 | Cost: 0.0123980 | Accuracy: 1.0000000 \n",
            "Iter:    20 | Cost: 0.0085416 | Accuracy: 1.0000000 \n",
            "Iter:    21 | Cost: 0.0053549 | Accuracy: 1.0000000 \n",
            "Iter:    22 | Cost: 0.0065759 | Accuracy: 1.0000000 \n",
            "Iter:    23 | Cost: 0.0024883 | Accuracy: 1.0000000 \n",
            "Iter:    24 | Cost: 0.0029102 | Accuracy: 1.0000000 \n",
            "Iter:    25 | Cost: 0.0023471 | Accuracy: 1.0000000 \n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# A real-world example\n",
        "\n",
        "## Iris Classification\n"
      ],
      "metadata": {
        "id": "hGpW6r-pYcmc"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Quantum and Classical Nodes\n",
        "\n",
        "To encode real-valued vectors into the amplitudes of a quantum state, we use a 2-qubit simulator"
      ],
      "metadata": {
        "id": "Q-XzYHsFYhKy"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "dev = qml.device(\"default.qubit\", wires=2)"
      ],
      "metadata": {
        "id": "QY4q5OkrYPW3"
      },
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "State preparation is not as simple as when we represent a bitstring with a basis state. Every input $x$ has to be translated into a set of angles which can get fed into a small routine for state preparation. To simplify things a bit, we will work with data from the positive subspace, so that we can ignore signs (which would require another cascade of rotations around the z axis).\n",
        "\n",
        "\n",
        "The circuit is coded according to the scheme in [ Möttönen, et al. (2004),](https://arxiv.org/abs/quant-ph/0407010). We had to decompose controlled Y-axis rotations into more basic circuits."
      ],
      "metadata": {
        "id": "TvwH185lYpFD"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def get_angles(x):\n",
        "\n",
        "    beta0 = 2 * np.arcsin(np.sqrt(x[1] ** 2) / np.sqrt(x[0] ** 2 + x[1] ** 2 + 1e-12))\n",
        "    beta1 = 2 * np.arcsin(np.sqrt(x[3] ** 2) / np.sqrt(x[2] ** 2 + x[3] ** 2 + 1e-12))\n",
        "    beta2 = 2 * np.arcsin(\n",
        "        np.sqrt(x[2] ** 2 + x[3] ** 2)\n",
        "        / np.sqrt(x[0] ** 2 + x[1] ** 2 + x[2] ** 2 + x[3] ** 2)\n",
        "    )\n",
        "\n",
        "    return np.array([beta2, -beta1 / 2, beta1 / 2, -beta0 / 2, beta0 / 2])\n",
        "\n",
        "\n",
        "def statepreparation(a):\n",
        "    qml.RY(a[0], wires=0)\n",
        "\n",
        "    qml.CNOT(wires=[0, 1])\n",
        "    qml.RY(a[1], wires=1)\n",
        "    qml.CNOT(wires=[0, 1])\n",
        "    qml.RY(a[2], wires=1)\n",
        "\n",
        "    qml.PauliX(wires=0)\n",
        "    qml.CNOT(wires=[0, 1])\n",
        "    qml.RY(a[3], wires=1)\n",
        "    qml.CNOT(wires=[0, 1])\n",
        "    qml.RY(a[4], wires=1)\n",
        "    qml.PauliX(wires=0)"
      ],
      "metadata": {
        "id": "VRVTDjaCYm3s"
      },
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Let's see if it works"
      ],
      "metadata": {
        "id": "tknuOFSWY8_F"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = np.array([0.53896774, 0.79503606, 0.27826503, 0.0], requires_grad=False)\n",
        "ang = get_angles(x)\n",
        "\n",
        "\n",
        "@qml.qnode(dev)\n",
        "def test(angles):\n",
        "\n",
        "    statepreparation(angles)\n",
        "\n",
        "    return qml.expval(qml.PauliZ(0))\n",
        "\n",
        "\n",
        "test(ang)\n",
        "\n",
        "print(\"x               : \", x)\n",
        "print(\"angles          : \", ang)\n",
        "print(\"amplitude vector: \", np.real(dev.state))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "19oA8dW0Y8Ch",
        "outputId": "a6bed5f3-aa3e-4c72-f390-f8a2e01eb224"
      },
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "x               :  [0.53896774 0.79503606 0.27826503 0.        ]\n",
            "angles          :  [ 0.56397465 -0.          0.         -0.97504604  0.97504604]\n",
            "amplitude vector:  [ 5.38967743e-01  7.95036065e-01  2.78265032e-01 -2.77555756e-17]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Note that the `default.qubit` simulator provides a shortcut to `statepreparation` with the command `qml.QubitStateVector(x, wires=[0, 1])`. However, some devices may not support an arbitrary state-preparation routine.\n",
        "\n",
        "Since we are working with only 2 qubits now, we need to update the layer function as well."
      ],
      "metadata": {
        "id": "ETTjKDPaZDL5"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def layer(W):\n",
        "    qml.Rot(W[0, 0], W[0, 1], W[0, 2], wires=0)\n",
        "    qml.Rot(W[1, 0], W[1, 1], W[1, 2], wires=1)\n",
        "    qml.CNOT(wires=[0, 1])"
      ],
      "metadata": {
        "id": "eDAIG1f9ZAaP"
      },
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "The variational classifier model and its cost remain essentially the same, but we have to reload them with the new state preparation and layer functions."
      ],
      "metadata": {
        "id": "RRldgSmxZNRA"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "@qml.qnode(dev)\n",
        "def circuit(weights, angles):\n",
        "    statepreparation(angles)\n",
        "\n",
        "    for W in weights:\n",
        "        layer(W)\n",
        "\n",
        "    return qml.expval(qml.PauliZ(0))\n",
        "\n",
        "\n",
        "def variational_classifier(weights, bias, angles):\n",
        "    return circuit(weights, angles) + bias\n",
        "\n",
        "\n",
        "def cost(weights, bias, features, labels):\n",
        "    predictions = [variational_classifier(weights, bias, f) for f in features]\n",
        "    return square_loss(labels, predictions)\n"
      ],
      "metadata": {
        "id": "iqU4t_beZLxL"
      },
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Data\n",
        "\n",
        "We then load the Iris data set. There is a bit of preprocessing to do in order to encode the inputs into the amplitudes of a quantum state. In the last preprocessing step, we translate the inputs x to rotation angles using the `get_angles` function we defined above.\n",
        "\n",
        "\n",
        "Get the data from [here](https://raw.githubusercontent.com/XanaduAI/qml/master/demonstrations/variational_classifier/data/iris_classes1and2_scaled.txt)"
      ],
      "metadata": {
        "id": "Xk7E5MgUZR8R"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "data = np.loadtxt(\"variational/data/iris.txt\")\n",
        "X = data[:, 0:2]\n",
        "print(\"First X sample (original)  :\", X[0])\n",
        "\n",
        "# pad the vectors to size 2^2 with constant values\n",
        "padding = 0.3 * np.ones((len(X), 1))\n",
        "X_pad = np.c_[np.c_[X, padding], np.zeros((len(X), 1))]\n",
        "print(\"First X sample (padded)    :\", X_pad[0])\n",
        "\n",
        "# normalize each input\n",
        "normalization = np.sqrt(np.sum(X_pad ** 2, -1))\n",
        "X_norm = (X_pad.T / normalization).T\n",
        "print(\"First X sample (normalized):\", X_norm[0])\n",
        "\n",
        "# angles for state preparation are new features\n",
        "features = np.array([get_angles(x) for x in X_norm], requires_grad=False)\n",
        "print(\"First features sample      :\", features[0])\n",
        "\n",
        "Y = data[:, -1]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "veSusQZ7ZQek",
        "outputId": "2f560ca4-7ad2-46b7-e14a-0e6141c45c9b"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "First X sample (original)  : [0.4  0.75]\n",
            "First X sample (padded)    : [0.4  0.75 0.3  0.  ]\n",
            "First X sample (normalized): [0.44376016 0.83205029 0.33282012 0.        ]\n",
            "First features sample      : [ 0.67858523 -0.          0.         -1.080839    1.080839  ]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "These angles are our new features, which is why we have renamed X to “features” above. Let’s plot the stages of preprocessing and play around with the dimensions (dim1, dim2). Some of them still separate the classes well, while others are less informative."
      ],
      "metadata": {
        "id": "cFYRjyW1Z0m2"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.figure()\n",
        "plt.scatter(X[:, 0][Y == 1], X[:, 1][Y == 1], c=\"b\", marker=\"o\", edgecolors=\"k\")\n",
        "plt.scatter(X[:, 0][Y == -1], X[:, 1][Y == -1], c=\"r\", marker=\"o\", edgecolors=\"k\")\n",
        "plt.title(\"Original data\")\n",
        "plt.show()\n",
        "\n",
        "plt.figure()\n",
        "dim1 = 0\n",
        "dim2 = 1\n",
        "plt.scatter(\n",
        "    X_norm[:, dim1][Y == 1], X_norm[:, dim2][Y == 1], c=\"b\", marker=\"o\", edgecolors=\"k\"\n",
        ")\n",
        "plt.scatter(\n",
        "    X_norm[:, dim1][Y == -1], X_norm[:, dim2][Y == -1], c=\"r\", marker=\"o\", edgecolors=\"k\"\n",
        ")\n",
        "plt.title(\"Padded and normalised data (dims {} and {})\".format(dim1, dim2))\n",
        "plt.show()\n",
        "\n",
        "plt.figure()\n",
        "dim1 = 0\n",
        "dim2 = 3\n",
        "plt.scatter(\n",
        "    features[:, dim1][Y == 1], features[:, dim2][Y == 1], c=\"b\", marker=\"o\", edgecolors=\"k\"\n",
        ")\n",
        "plt.scatter(\n",
        "    features[:, dim1][Y == -1], features[:, dim2][Y == -1], c=\"r\", marker=\"o\", edgecolors=\"k\"\n",
        ")\n",
        "plt.title(\"Feature vectors (dims {} and {})\".format(dim1, dim2))\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 809
        },
        "id": "LOJeLTvXZvDB",
        "outputId": "be8d1c92-354f-42ad-d8e5-b2a1b08887d3"
      },
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "This time we want to generalize from the data samples. To monitor the generalization performance, the data is split into training and validation set."
      ],
      "metadata": {
        "id": "IFi-ZNqjZ6u2"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "np.random.seed(0)\n",
        "num_data = len(Y)\n",
        "num_train = int(0.75 * num_data)\n",
        "index = np.random.permutation(range(num_data))\n",
        "feats_train = features[index[:num_train]]\n",
        "Y_train = Y[index[:num_train]]\n",
        "feats_val = features[index[num_train:]]\n",
        "Y_val = Y[index[num_train:]]\n",
        "\n",
        "# We need these later for plotting\n",
        "X_train = X[index[:num_train]]\n",
        "X_val = X[index[num_train:]]\n"
      ],
      "metadata": {
        "id": "tnWL8fkmZ3xe"
      },
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Optimization\n",
        "\n",
        "First we initialize the variables\n"
      ],
      "metadata": {
        "id": "7FLHOvgfZ97Z"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "num_qubits = 2\n",
        "num_layers = 6\n",
        "\n",
        "weights_init = 0.01 * np.random.randn(num_layers, num_qubits, 3, requires_grad=True)\n",
        "bias_init = np.array(0.0, requires_grad=True)"
      ],
      "metadata": {
        "id": "J3bNelPXZ9HX"
      },
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Again, we optimize the cost"
      ],
      "metadata": {
        "id": "qmfW_lEjaEhZ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "opt = NesterovMomentumOptimizer(0.01)\n",
        "batch_size = 5\n",
        "\n",
        "# train the variational classifier\n",
        "weights = weights_init\n",
        "bias = bias_init\n",
        "for it in range(60):\n",
        "\n",
        "    # Update the weights by one optimizer step\n",
        "    batch_index = np.random.randint(0, num_train, (batch_size,))\n",
        "    feats_train_batch = feats_train[batch_index]\n",
        "    Y_train_batch = Y_train[batch_index]\n",
        "    weights, bias, _, _ = opt.step(cost, weights, bias, feats_train_batch, Y_train_batch)\n",
        "\n",
        "    # Compute predictions on train and validation set\n",
        "    predictions_train = [np.sign(variational_classifier(weights, bias, f)) for f in feats_train]\n",
        "    predictions_val = [np.sign(variational_classifier(weights, bias, f)) for f in feats_val]\n",
        "\n",
        "    # Compute accuracy on train and validation set\n",
        "    acc_train = accuracy(Y_train, predictions_train)\n",
        "    acc_val = accuracy(Y_val, predictions_val)\n",
        "\n",
        "    print(\n",
        "        \"Iter: {:5d} | Cost: {:0.7f} | Acc train: {:0.7f} | Acc validation: {:0.7f} \"\n",
        "        \"\".format(it + 1, cost(weights, bias, features, Y), acc_train, acc_val)\n",
        "    )"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rQlV4uryaDwO",
        "outputId": "150d6796-9cc4-4dbe-daa2-58800fa5d75b"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Iter:     1 | Cost: 1.4490948 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     2 | Cost: 1.3312057 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     3 | Cost: 1.1589332 | Acc train: 0.4533333 | Acc validation: 0.5600000 \n",
            "Iter:     4 | Cost: 0.9806934 | Acc train: 0.4800000 | Acc validation: 0.5600000 \n",
            "Iter:     5 | Cost: 0.8865623 | Acc train: 0.6133333 | Acc validation: 0.7600000 \n",
            "Iter:     6 | Cost: 0.8580769 | Acc train: 0.6933333 | Acc validation: 0.7600000 \n",
            "Iter:     7 | Cost: 0.8473132 | Acc train: 0.7200000 | Acc validation: 0.6800000 \n",
            "Iter:     8 | Cost: 0.8177533 | Acc train: 0.7333333 | Acc validation: 0.6800000 \n",
            "Iter:     9 | Cost: 0.8001100 | Acc train: 0.7466667 | Acc validation: 0.6800000 \n",
            "Iter:    10 | Cost: 0.7681053 | Acc train: 0.8000000 | Acc validation: 0.7600000 \n",
            "Iter:    11 | Cost: 0.7440015 | Acc train: 0.8133333 | Acc validation: 0.9600000 \n",
            "Iter:    12 | Cost: 0.7583777 | Acc train: 0.6800000 | Acc validation: 0.7600000 \n",
            "Iter:    13 | Cost: 0.7896372 | Acc train: 0.6533333 | Acc validation: 0.7200000 \n",
            "Iter:    14 | Cost: 0.8397790 | Acc train: 0.6133333 | Acc validation: 0.6400000 \n",
            "Iter:    15 | Cost: 0.8632423 | Acc train: 0.5733333 | Acc validation: 0.6000000 \n",
            "Iter:    16 | Cost: 0.8693517 | Acc train: 0.5733333 | Acc validation: 0.6000000 \n",
            "Iter:    17 | Cost: 0.8350625 | Acc train: 0.6266667 | Acc validation: 0.6400000 \n",
            "Iter:    18 | Cost: 0.7966558 | Acc train: 0.6266667 | Acc validation: 0.6400000 \n",
            "Iter:    19 | Cost: 0.7563381 | Acc train: 0.6800000 | Acc validation: 0.7600000 \n",
            "Iter:    20 | Cost: 0.7315459 | Acc train: 0.7733333 | Acc validation: 0.8000000 \n",
            "Iter:    21 | Cost: 0.7182359 | Acc train: 0.8800000 | Acc validation: 0.9600000 \n",
            "Iter:    22 | Cost: 0.7339132 | Acc train: 0.8133333 | Acc validation: 0.7200000 \n",
            "Iter:    23 | Cost: 0.7498571 | Acc train: 0.6533333 | Acc validation: 0.6400000 \n",
            "Iter:    24 | Cost: 0.7593763 | Acc train: 0.6000000 | Acc validation: 0.6400000 \n",
            "Iter:    25 | Cost: 0.7423832 | Acc train: 0.6800000 | Acc validation: 0.6400000 \n",
            "Iter:    26 | Cost: 0.7361186 | Acc train: 0.7333333 | Acc validation: 0.6800000 \n",
            "Iter:    27 | Cost: 0.7300201 | Acc train: 0.7600000 | Acc validation: 0.6800000 \n",
            "Iter:    28 | Cost: 0.7360923 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    29 | Cost: 0.7371325 | Acc train: 0.6533333 | Acc validation: 0.6400000 \n",
            "Iter:    30 | Cost: 0.7234520 | Acc train: 0.7466667 | Acc validation: 0.6800000 \n",
            "Iter:    31 | Cost: 0.7112089 | Acc train: 0.8133333 | Acc validation: 0.7600000 \n",
            "Iter:    32 | Cost: 0.6923495 | Acc train: 0.9200000 | Acc validation: 0.9200000 \n",
            "Iter:    33 | Cost: 0.6792510 | Acc train: 0.9200000 | Acc validation: 1.0000000 \n",
            "Iter:    34 | Cost: 0.6756795 | Acc train: 0.9200000 | Acc validation: 0.9600000 \n",
            "Iter:    35 | Cost: 0.6787274 | Acc train: 0.8933333 | Acc validation: 0.8000000 \n",
            "Iter:    36 | Cost: 0.6938181 | Acc train: 0.7466667 | Acc validation: 0.6800000 \n",
            "Iter:    37 | Cost: 0.6765225 | Acc train: 0.8266667 | Acc validation: 0.8000000 \n",
            "Iter:    38 | Cost: 0.6708057 | Acc train: 0.8266667 | Acc validation: 0.8000000 \n",
            "Iter:    39 | Cost: 0.6740248 | Acc train: 0.7600000 | Acc validation: 0.7200000 \n",
            "Iter:    40 | Cost: 0.6835014 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    41 | Cost: 0.6686811 | Acc train: 0.7333333 | Acc validation: 0.7200000 \n",
            "Iter:    42 | Cost: 0.6217025 | Acc train: 0.8933333 | Acc validation: 0.8400000 \n",
            "Iter:    43 | Cost: 0.6004307 | Acc train: 0.9066667 | Acc validation: 0.9200000 \n",
            "Iter:    44 | Cost: 0.5910768 | Acc train: 0.9066667 | Acc validation: 0.9200000 \n",
            "Iter:    45 | Cost: 0.5694770 | Acc train: 0.9200000 | Acc validation: 0.9600000 \n",
            "Iter:    46 | Cost: 0.5588385 | Acc train: 0.9200000 | Acc validation: 0.9600000 \n",
            "Iter:    47 | Cost: 0.5381473 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    48 | Cost: 0.5205658 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    49 | Cost: 0.5064983 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    50 | Cost: 0.4969733 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    51 | Cost: 0.4782257 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    52 | Cost: 0.4536953 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    53 | Cost: 0.4350300 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    54 | Cost: 0.4272909 | Acc train: 0.9733333 | Acc validation: 0.9600000 \n",
            "Iter:    55 | Cost: 0.4288929 | Acc train: 0.9466667 | Acc validation: 0.9200000 \n",
            "Iter:    56 | Cost: 0.4261037 | Acc train: 0.9333333 | Acc validation: 0.9200000 \n",
            "Iter:    57 | Cost: 0.4082663 | Acc train: 0.9466667 | Acc validation: 0.9200000 \n",
            "Iter:    58 | Cost: 0.3698736 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    59 | Cost: 0.3420686 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    60 | Cost: 0.3253480 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:     1 | Cost: 1.4343058 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     2 | Cost: 1.2823996 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     3 | Cost: 1.1375452 | Acc train: 0.4533333 | Acc validation: 0.5600000 \n",
            "Iter:     4 | Cost: 1.0266429 | Acc train: 0.4666667 | Acc validation: 0.5600000 \n",
            "Iter:     5 | Cost: 0.9357324 | Acc train: 0.5200000 | Acc validation: 0.6000000 \n",
            "Iter:     6 | Cost: 0.8728969 | Acc train: 0.5866667 | Acc validation: 0.7600000 \n",
            "Iter:     7 | Cost: 0.8375035 | Acc train: 0.6800000 | Acc validation: 0.7600000 \n",
            "Iter:     8 | Cost: 0.8086149 | Acc train: 0.7600000 | Acc validation: 0.8000000 \n",
            "Iter:     9 | Cost: 0.7920516 | Acc train: 0.8000000 | Acc validation: 0.8000000 \n",
            "Iter:    10 | Cost: 0.7786248 | Acc train: 0.7866667 | Acc validation: 0.7600000 \n",
            "Iter:    11 | Cost: 0.7787500 | Acc train: 0.7466667 | Acc validation: 0.6800000 \n",
            "Iter:    12 | Cost: 0.7684224 | Acc train: 0.7333333 | Acc validation: 0.6800000 \n",
            "Iter:    13 | Cost: 0.7468000 | Acc train: 0.7733333 | Acc validation: 0.7200000 \n",
            "Iter:    14 | Cost: 0.7421490 | Acc train: 0.7600000 | Acc validation: 0.6800000 \n",
            "Iter:    15 | Cost: 0.7303465 | Acc train: 0.8400000 | Acc validation: 0.8400000 \n",
            "Iter:    16 | Cost: 0.7218077 | Acc train: 0.8800000 | Acc validation: 0.9600000 \n",
            "Iter:    17 | Cost: 0.7258294 | Acc train: 0.8133333 | Acc validation: 0.9200000 \n",
            "Iter:    18 | Cost: 0.7326151 | Acc train: 0.8266667 | Acc validation: 0.9200000 \n",
            "Iter:    19 | Cost: 0.7319506 | Acc train: 0.8133333 | Acc validation: 0.9200000 \n",
            "Iter:    20 | Cost: 0.7279744 | Acc train: 0.8266667 | Acc validation: 0.9200000 \n",
            "Iter:    21 | Cost: 0.7210869 | Acc train: 0.8800000 | Acc validation: 0.9600000 \n",
            "Iter:    22 | Cost: 0.7155283 | Acc train: 0.8800000 | Acc validation: 0.9600000 \n",
            "Iter:    23 | Cost: 0.7145997 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    24 | Cost: 0.7093569 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    25 | Cost: 0.7088305 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    26 | Cost: 0.6993786 | Acc train: 0.8666667 | Acc validation: 0.9200000 \n",
            "Iter:    27 | Cost: 0.6911881 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    28 | Cost: 0.6889397 | Acc train: 0.8666667 | Acc validation: 0.9200000 \n",
            "Iter:    29 | Cost: 0.6753967 | Acc train: 0.9066667 | Acc validation: 0.9600000 \n",
            "Iter:    30 | Cost: 0.6664242 | Acc train: 0.9200000 | Acc validation: 1.0000000 \n",
            "Iter:    31 | Cost: 0.6723145 | Acc train: 0.8533333 | Acc validation: 0.8000000 \n",
            "Iter:    32 | Cost: 0.6931079 | Acc train: 0.7200000 | Acc validation: 0.7200000 \n",
            "Iter:    33 | Cost: 0.7069079 | Acc train: 0.6000000 | Acc validation: 0.5600000 \n",
            "Iter:    34 | Cost: 0.7181192 | Acc train: 0.5866667 | Acc validation: 0.5200000 \n",
            "Iter:    35 | Cost: 0.6732392 | Acc train: 0.7333333 | Acc validation: 0.7600000 \n",
            "Iter:    36 | Cost: 0.6365898 | Acc train: 0.8800000 | Acc validation: 0.8400000 \n",
            "Iter:    37 | Cost: 0.6135017 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    38 | Cost: 0.6173204 | Acc train: 0.9200000 | Acc validation: 0.9200000 \n",
            "Iter:    39 | Cost: 0.6181296 | Acc train: 0.8800000 | Acc validation: 0.9200000 \n",
            "Iter:    40 | Cost: 0.6116891 | Acc train: 0.8933333 | Acc validation: 0.9200000 \n",
            "Iter:    41 | Cost: 0.6185524 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    42 | Cost: 0.6523419 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    43 | Cost: 0.6679560 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    44 | Cost: 0.6674770 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    45 | Cost: 0.6545126 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    46 | Cost: 0.6160590 | Acc train: 0.7600000 | Acc validation: 0.7200000 \n",
            "Iter:    47 | Cost: 0.5706839 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    48 | Cost: 0.5333502 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    49 | Cost: 0.5102810 | Acc train: 0.9733333 | Acc validation: 0.9600000 \n",
            "Iter:    50 | Cost: 0.4852555 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    51 | Cost: 0.4586707 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    52 | Cost: 0.4386251 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    53 | Cost: 0.4230347 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    54 | Cost: 0.4083736 | Acc train: 0.9733333 | Acc validation: 1.0000000 \n",
            "Iter:    55 | Cost: 0.3942317 | Acc train: 0.9733333 | Acc validation: 1.0000000 \n",
            "Iter:    56 | Cost: 0.3775060 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    57 | Cost: 0.3639581 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    58 | Cost: 0.3553534 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    59 | Cost: 0.3466206 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    60 | Cost: 0.3390452 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:     1 | Cost: 1.4255599 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     2 | Cost: 1.3438758 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     3 | Cost: 1.2629339 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     4 | Cost: 1.1787994 | Acc train: 0.4800000 | Acc validation: 0.5600000 \n",
            "Iter:     5 | Cost: 1.0637003 | Acc train: 0.4800000 | Acc validation: 0.5600000 \n",
            "Iter:     6 | Cost: 0.9474419 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     7 | Cost: 0.8432361 | Acc train: 0.6266667 | Acc validation: 0.7600000 \n",
            "Iter:     8 | Cost: 0.8053478 | Acc train: 0.7600000 | Acc validation: 0.8400000 \n",
            "Iter:     9 | Cost: 0.8343735 | Acc train: 0.6400000 | Acc validation: 0.6400000 \n",
            "Iter:    10 | Cost: 0.8627575 | Acc train: 0.5600000 | Acc validation: 0.4800000 \n",
            "Iter:    11 | Cost: 0.8758388 | Acc train: 0.5333333 | Acc validation: 0.4400000 \n",
            "Iter:    12 | Cost: 0.8380779 | Acc train: 0.5333333 | Acc validation: 0.4800000 \n",
            "Iter:    13 | Cost: 0.7548203 | Acc train: 0.7200000 | Acc validation: 0.6800000 \n",
            "Iter:    14 | Cost: 0.7201367 | Acc train: 0.8800000 | Acc validation: 1.0000000 \n",
            "Iter:    15 | Cost: 0.7411357 | Acc train: 0.7466667 | Acc validation: 0.7600000 \n",
            "Iter:    16 | Cost: 0.8108804 | Acc train: 0.6400000 | Acc validation: 0.6400000 \n",
            "Iter:    17 | Cost: 0.8864019 | Acc train: 0.5733333 | Acc validation: 0.6000000 \n",
            "Iter:    18 | Cost: 0.9668866 | Acc train: 0.5466667 | Acc validation: 0.5600000 \n",
            "Iter:    19 | Cost: 1.0392077 | Acc train: 0.5466667 | Acc validation: 0.5600000 \n",
            "Iter:    20 | Cost: 1.0797906 | Acc train: 0.5466667 | Acc validation: 0.5600000 \n",
            "Iter:    21 | Cost: 1.0896552 | Acc train: 0.5466667 | Acc validation: 0.5600000 \n",
            "Iter:    22 | Cost: 1.0510823 | Acc train: 0.5466667 | Acc validation: 0.5600000 \n",
            "Iter:    23 | Cost: 0.9974279 | Acc train: 0.5466667 | Acc validation: 0.5600000 \n",
            "Iter:    24 | Cost: 0.9336845 | Acc train: 0.5600000 | Acc validation: 0.6000000 \n",
            "Iter:    25 | Cost: 0.8960334 | Acc train: 0.5600000 | Acc validation: 0.6000000 \n",
            "Iter:    26 | Cost: 0.8213553 | Acc train: 0.6133333 | Acc validation: 0.6000000 \n",
            "Iter:    27 | Cost: 0.7364379 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    28 | Cost: 0.6634673 | Acc train: 0.8133333 | Acc validation: 0.8000000 \n",
            "Iter:    29 | Cost: 0.6376125 | Acc train: 0.8666667 | Acc validation: 0.9200000 \n",
            "Iter:    30 | Cost: 0.6493091 | Acc train: 0.9066667 | Acc validation: 0.9600000 \n",
            "Iter:    31 | Cost: 0.7024020 | Acc train: 0.7200000 | Acc validation: 0.6800000 \n",
            "Iter:    32 | Cost: 0.7303731 | Acc train: 0.6400000 | Acc validation: 0.6400000 \n",
            "Iter:    33 | Cost: 0.6950323 | Acc train: 0.6933333 | Acc validation: 0.6400000 \n",
            "Iter:    34 | Cost: 0.6417225 | Acc train: 0.7733333 | Acc validation: 0.7200000 \n",
            "Iter:    35 | Cost: 0.5857049 | Acc train: 0.9066667 | Acc validation: 0.9200000 \n",
            "Iter:    36 | Cost: 0.5349393 | Acc train: 0.9333333 | Acc validation: 1.0000000 \n",
            "Iter:    37 | Cost: 0.4970762 | Acc train: 0.9200000 | Acc validation: 1.0000000 \n",
            "Iter:    38 | Cost: 0.4692431 | Acc train: 0.9600000 | Acc validation: 0.9600000 \n",
            "Iter:    39 | Cost: 0.4483889 | Acc train: 0.9466667 | Acc validation: 0.9200000 \n",
            "Iter:    40 | Cost: 0.4300282 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    41 | Cost: 0.4171287 | Acc train: 0.9466667 | Acc validation: 0.9200000 \n",
            "Iter:    42 | Cost: 0.4040047 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    43 | Cost: 0.3927142 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    44 | Cost: 0.3700163 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    45 | Cost: 0.3429887 | Acc train: 0.9733333 | Acc validation: 0.9600000 \n",
            "Iter:    46 | Cost: 0.3195273 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    47 | Cost: 0.3060264 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    48 | Cost: 0.2991315 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    49 | Cost: 0.3012888 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    50 | Cost: 0.3010054 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    51 | Cost: 0.2932498 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    52 | Cost: 0.2801224 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    53 | Cost: 0.2722974 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    54 | Cost: 0.2652886 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    55 | Cost: 0.2585635 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    56 | Cost: 0.2526253 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    57 | Cost: 0.2488337 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    58 | Cost: 0.2465852 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    59 | Cost: 0.2451489 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    60 | Cost: 0.2452397 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:     1 | Cost: 1.4508493 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     2 | Cost: 1.3327317 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     3 | Cost: 1.2106946 | Acc train: 0.4800000 | Acc validation: 0.5600000 \n",
            "Iter:     4 | Cost: 1.0914785 | Acc train: 0.4666667 | Acc validation: 0.5600000 \n",
            "Iter:     5 | Cost: 0.9903260 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     6 | Cost: 0.8694936 | Acc train: 0.5733333 | Acc validation: 0.7600000 \n",
            "Iter:     7 | Cost: 0.8172001 | Acc train: 0.7333333 | Acc validation: 0.7200000 \n",
            "Iter:     8 | Cost: 0.8197257 | Acc train: 0.7333333 | Acc validation: 0.6800000 \n",
            "Iter:     9 | Cost: 0.8211951 | Acc train: 0.6400000 | Acc validation: 0.6400000 \n",
            "Iter:    10 | Cost: 0.8273930 | Acc train: 0.6133333 | Acc validation: 0.6400000 \n",
            "Iter:    11 | Cost: 0.7877739 | Acc train: 0.6400000 | Acc validation: 0.6400000 \n",
            "Iter:    12 | Cost: 0.7529707 | Acc train: 0.7466667 | Acc validation: 0.6800000 \n",
            "Iter:    13 | Cost: 0.7291275 | Acc train: 0.8400000 | Acc validation: 0.8800000 \n",
            "Iter:    14 | Cost: 0.7214400 | Acc train: 0.8800000 | Acc validation: 1.0000000 \n",
            "Iter:    15 | Cost: 0.7229250 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    16 | Cost: 0.7326839 | Acc train: 0.8000000 | Acc validation: 0.9200000 \n",
            "Iter:    17 | Cost: 0.7383741 | Acc train: 0.8000000 | Acc validation: 0.8000000 \n",
            "Iter:    18 | Cost: 0.7355361 | Acc train: 0.8000000 | Acc validation: 0.9200000 \n",
            "Iter:    19 | Cost: 0.7326418 | Acc train: 0.8266667 | Acc validation: 0.9200000 \n",
            "Iter:    20 | Cost: 0.7340785 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    21 | Cost: 0.7357967 | Acc train: 0.8933333 | Acc validation: 1.0000000 \n",
            "Iter:    22 | Cost: 0.7647163 | Acc train: 0.5333333 | Acc validation: 0.4400000 \n",
            "Iter:    23 | Cost: 0.7819105 | Acc train: 0.5333333 | Acc validation: 0.4400000 \n",
            "Iter:    24 | Cost: 0.7741221 | Acc train: 0.5333333 | Acc validation: 0.4400000 \n",
            "Iter:    25 | Cost: 0.7512644 | Acc train: 0.5600000 | Acc validation: 0.4400000 \n",
            "Iter:    26 | Cost: 0.7230790 | Acc train: 0.8666667 | Acc validation: 0.8400000 \n",
            "Iter:    27 | Cost: 0.7042868 | Acc train: 0.9066667 | Acc validation: 0.9600000 \n",
            "Iter:    28 | Cost: 0.7111816 | Acc train: 0.8266667 | Acc validation: 0.9200000 \n",
            "Iter:    29 | Cost: 0.7099015 | Acc train: 0.7733333 | Acc validation: 0.7600000 \n",
            "Iter:    30 | Cost: 0.7237095 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    31 | Cost: 0.7197422 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    32 | Cost: 0.7029253 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    33 | Cost: 0.6729647 | Acc train: 0.7733333 | Acc validation: 0.7600000 \n",
            "Iter:    34 | Cost: 0.6515481 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    35 | Cost: 0.6356539 | Acc train: 0.9066667 | Acc validation: 1.0000000 \n",
            "Iter:    36 | Cost: 0.6461299 | Acc train: 0.8933333 | Acc validation: 0.8400000 \n",
            "Iter:    37 | Cost: 0.6687867 | Acc train: 0.7733333 | Acc validation: 0.7600000 \n",
            "Iter:    38 | Cost: 0.6883987 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    39 | Cost: 0.6894650 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    40 | Cost: 0.6590391 | Acc train: 0.7333333 | Acc validation: 0.7600000 \n",
            "Iter:    41 | Cost: 0.6388764 | Acc train: 0.8266667 | Acc validation: 0.8000000 \n",
            "Iter:    42 | Cost: 0.6276480 | Acc train: 0.8266667 | Acc validation: 0.8000000 \n",
            "Iter:    43 | Cost: 0.6092665 | Acc train: 0.8400000 | Acc validation: 0.8000000 \n",
            "Iter:    44 | Cost: 0.5618256 | Acc train: 0.9333333 | Acc validation: 1.0000000 \n",
            "Iter:    45 | Cost: 0.5385674 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    46 | Cost: 0.5290338 | Acc train: 0.9333333 | Acc validation: 0.9200000 \n",
            "Iter:    47 | Cost: 0.5201610 | Acc train: 0.9333333 | Acc validation: 0.9200000 \n",
            "Iter:    48 | Cost: 0.5117306 | Acc train: 0.8933333 | Acc validation: 0.9200000 \n",
            "Iter:    49 | Cost: 0.5096188 | Acc train: 0.8800000 | Acc validation: 0.9200000 \n",
            "Iter:    50 | Cost: 0.5037851 | Acc train: 0.8666667 | Acc validation: 0.9200000 \n",
            "Iter:    51 | Cost: 0.4861966 | Acc train: 0.8800000 | Acc validation: 0.9200000 \n",
            "Iter:    52 | Cost: 0.4640838 | Acc train: 0.8933333 | Acc validation: 0.9200000 \n",
            "Iter:    53 | Cost: 0.4597279 | Acc train: 0.8933333 | Acc validation: 0.9200000 \n",
            "Iter:    54 | Cost: 0.4502758 | Acc train: 0.8933333 | Acc validation: 0.9200000 \n",
            "Iter:    55 | Cost: 0.4431786 | Acc train: 0.8800000 | Acc validation: 0.9200000 \n",
            "Iter:    56 | Cost: 0.4155888 | Acc train: 0.9333333 | Acc validation: 0.9200000 \n",
            "Iter:    57 | Cost: 0.3905666 | Acc train: 0.9466667 | Acc validation: 0.9200000 \n",
            "Iter:    58 | Cost: 0.3745334 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    59 | Cost: 0.3528154 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    60 | Cost: 0.3352920 | Acc train: 0.9733333 | Acc validation: 0.9600000 \n",
            "Iter:     1 | Cost: 1.4817982 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     2 | Cost: 1.3786181 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     3 | Cost: 1.2652595 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     4 | Cost: 1.1304112 | Acc train: 0.4800000 | Acc validation: 0.5600000 \n",
            "Iter:     5 | Cost: 1.0085131 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     6 | Cost: 0.9046338 | Acc train: 0.5466667 | Acc validation: 0.6400000 \n",
            "Iter:     7 | Cost: 0.8427924 | Acc train: 0.6400000 | Acc validation: 0.7600000 \n",
            "Iter:     8 | Cost: 0.8188130 | Acc train: 0.7200000 | Acc validation: 0.7600000 \n",
            "Iter:     9 | Cost: 0.8488776 | Acc train: 0.6400000 | Acc validation: 0.6400000 \n",
            "Iter:    10 | Cost: 0.8631006 | Acc train: 0.5866667 | Acc validation: 0.5600000 \n",
            "Iter:    11 | Cost: 0.8446128 | Acc train: 0.5866667 | Acc validation: 0.5600000 \n",
            "Iter:    12 | Cost: 0.8179604 | Acc train: 0.6266667 | Acc validation: 0.6400000 \n",
            "Iter:    13 | Cost: 0.7471253 | Acc train: 0.8266667 | Acc validation: 0.7600000 \n",
            "Iter:    14 | Cost: 0.7236662 | Acc train: 0.8133333 | Acc validation: 0.9200000 \n",
            "Iter:    15 | Cost: 0.7531339 | Acc train: 0.6800000 | Acc validation: 0.7600000 \n",
            "Iter:    16 | Cost: 0.8199139 | Acc train: 0.6400000 | Acc validation: 0.6400000 \n",
            "Iter:    17 | Cost: 0.8576253 | Acc train: 0.6000000 | Acc validation: 0.6000000 \n",
            "Iter:    18 | Cost: 0.8620869 | Acc train: 0.6133333 | Acc validation: 0.6000000 \n",
            "Iter:    19 | Cost: 0.8922519 | Acc train: 0.5733333 | Acc validation: 0.6000000 \n",
            "Iter:    20 | Cost: 0.8914904 | Acc train: 0.5733333 | Acc validation: 0.6000000 \n",
            "Iter:    21 | Cost: 0.8746611 | Acc train: 0.5866667 | Acc validation: 0.6000000 \n",
            "Iter:    22 | Cost: 0.8475210 | Acc train: 0.6133333 | Acc validation: 0.6000000 \n",
            "Iter:    23 | Cost: 0.8081607 | Acc train: 0.6400000 | Acc validation: 0.6400000 \n",
            "Iter:    24 | Cost: 0.7489328 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    25 | Cost: 0.7022298 | Acc train: 0.8266667 | Acc validation: 0.9200000 \n",
            "Iter:    26 | Cost: 0.6830616 | Acc train: 0.8933333 | Acc validation: 0.9600000 \n",
            "Iter:    27 | Cost: 0.7002880 | Acc train: 0.8133333 | Acc validation: 0.7600000 \n",
            "Iter:    28 | Cost: 0.7630780 | Acc train: 0.5333333 | Acc validation: 0.4400000 \n",
            "Iter:    29 | Cost: 0.7872350 | Acc train: 0.5200000 | Acc validation: 0.4400000 \n",
            "Iter:    30 | Cost: 0.7695144 | Acc train: 0.5200000 | Acc validation: 0.4400000 \n",
            "Iter:    31 | Cost: 0.7217459 | Acc train: 0.6266667 | Acc validation: 0.6400000 \n",
            "Iter:    32 | Cost: 0.6668024 | Acc train: 0.8133333 | Acc validation: 0.7600000 \n",
            "Iter:    33 | Cost: 0.6305481 | Acc train: 0.9066667 | Acc validation: 0.9600000 \n",
            "Iter:    34 | Cost: 0.5896638 | Acc train: 0.9066667 | Acc validation: 0.9600000 \n",
            "Iter:    35 | Cost: 0.5751247 | Acc train: 0.8666667 | Acc validation: 0.9200000 \n",
            "Iter:    36 | Cost: 0.5796049 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    37 | Cost: 0.5767425 | Acc train: 0.8000000 | Acc validation: 0.8400000 \n",
            "Iter:    38 | Cost: 0.5552557 | Acc train: 0.8133333 | Acc validation: 0.8400000 \n",
            "Iter:    39 | Cost: 0.5272815 | Acc train: 0.8400000 | Acc validation: 0.8800000 \n",
            "Iter:    40 | Cost: 0.4932512 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    41 | Cost: 0.4562280 | Acc train: 0.8933333 | Acc validation: 0.9200000 \n",
            "Iter:    42 | Cost: 0.4210985 | Acc train: 0.9466667 | Acc validation: 0.9200000 \n",
            "Iter:    43 | Cost: 0.3960916 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    44 | Cost: 0.3672423 | Acc train: 0.9733333 | Acc validation: 0.9600000 \n",
            "Iter:    45 | Cost: 0.3439843 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    46 | Cost: 0.3286987 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    47 | Cost: 0.3172685 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    48 | Cost: 0.3117586 | Acc train: 0.9733333 | Acc validation: 1.0000000 \n",
            "Iter:    49 | Cost: 0.3116143 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    50 | Cost: 0.3120069 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    51 | Cost: 0.3196927 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    52 | Cost: 0.3204193 | Acc train: 0.9333333 | Acc validation: 0.9600000 \n",
            "Iter:    53 | Cost: 0.3180826 | Acc train: 0.9333333 | Acc validation: 0.9600000 \n",
            "Iter:    54 | Cost: 0.3057767 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    55 | Cost: 0.2894456 | Acc train: 0.9733333 | Acc validation: 1.0000000 \n",
            "Iter:    56 | Cost: 0.2699037 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    57 | Cost: 0.2615616 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    58 | Cost: 0.2532180 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    59 | Cost: 0.2475917 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    60 | Cost: 0.2457421 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:     1 | Cost: 1.4808708 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     2 | Cost: 1.3880179 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     3 | Cost: 1.2248259 | Acc train: 0.4800000 | Acc validation: 0.5600000 \n",
            "Iter:     4 | Cost: 1.0775743 | Acc train: 0.4666667 | Acc validation: 0.5600000 \n",
            "Iter:     5 | Cost: 0.9647386 | Acc train: 0.5066667 | Acc validation: 0.6000000 \n",
            "Iter:     6 | Cost: 0.8878079 | Acc train: 0.6000000 | Acc validation: 0.7600000 \n",
            "Iter:     7 | Cost: 0.8583608 | Acc train: 0.7200000 | Acc validation: 0.6800000 \n",
            "Iter:     8 | Cost: 0.8426148 | Acc train: 0.7200000 | Acc validation: 0.6800000 \n",
            "Iter:     9 | Cost: 0.8294607 | Acc train: 0.7200000 | Acc validation: 0.6800000 \n",
            "Iter:    10 | Cost: 0.8192948 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    11 | Cost: 0.7828099 | Acc train: 0.7333333 | Acc validation: 0.6800000 \n",
            "Iter:    12 | Cost: 0.7458209 | Acc train: 0.8400000 | Acc validation: 0.8400000 \n",
            "Iter:    13 | Cost: 0.7256130 | Acc train: 0.8666667 | Acc validation: 1.0000000 \n",
            "Iter:    14 | Cost: 0.7178968 | Acc train: 0.8133333 | Acc validation: 0.9200000 \n",
            "Iter:    15 | Cost: 0.7185809 | Acc train: 0.8133333 | Acc validation: 0.9200000 \n",
            "Iter:    16 | Cost: 0.7160635 | Acc train: 0.8133333 | Acc validation: 0.9200000 \n",
            "Iter:    17 | Cost: 0.7164378 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    18 | Cost: 0.7223629 | Acc train: 0.8266667 | Acc validation: 0.9200000 \n",
            "Iter:    19 | Cost: 0.7295600 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    20 | Cost: 0.7297890 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    21 | Cost: 0.7230080 | Acc train: 0.8933333 | Acc validation: 1.0000000 \n",
            "Iter:    22 | Cost: 0.7214009 | Acc train: 0.8933333 | Acc validation: 1.0000000 \n",
            "Iter:    23 | Cost: 0.7177813 | Acc train: 0.8933333 | Acc validation: 1.0000000 \n",
            "Iter:    24 | Cost: 0.7135414 | Acc train: 0.9066667 | Acc validation: 0.9200000 \n",
            "Iter:    25 | Cost: 0.7095884 | Acc train: 0.8666667 | Acc validation: 0.8400000 \n",
            "Iter:    26 | Cost: 0.6990733 | Acc train: 0.8933333 | Acc validation: 0.8400000 \n",
            "Iter:    27 | Cost: 0.6880609 | Acc train: 0.9333333 | Acc validation: 0.9600000 \n",
            "Iter:    28 | Cost: 0.6793758 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    29 | Cost: 0.6709811 | Acc train: 0.9066667 | Acc validation: 0.8800000 \n",
            "Iter:    30 | Cost: 0.6776095 | Acc train: 0.7733333 | Acc validation: 0.8000000 \n",
            "Iter:    31 | Cost: 0.6931316 | Acc train: 0.6000000 | Acc validation: 0.5600000 \n",
            "Iter:    32 | Cost: 0.7238447 | Acc train: 0.5733333 | Acc validation: 0.5200000 \n",
            "Iter:    33 | Cost: 0.7222733 | Acc train: 0.5733333 | Acc validation: 0.5200000 \n",
            "Iter:    34 | Cost: 0.6791196 | Acc train: 0.6000000 | Acc validation: 0.5600000 \n",
            "Iter:    35 | Cost: 0.6203766 | Acc train: 0.7866667 | Acc validation: 0.8000000 \n",
            "Iter:    36 | Cost: 0.5827323 | Acc train: 0.8800000 | Acc validation: 0.8400000 \n",
            "Iter:    37 | Cost: 0.5520570 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    38 | Cost: 0.5442829 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    39 | Cost: 0.5364628 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    40 | Cost: 0.5277549 | Acc train: 0.9733333 | Acc validation: 0.9600000 \n",
            "Iter:    41 | Cost: 0.5124712 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    42 | Cost: 0.5052295 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    43 | Cost: 0.4797039 | Acc train: 0.9600000 | Acc validation: 0.9200000 \n",
            "Iter:    44 | Cost: 0.4422284 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    45 | Cost: 0.3973400 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    46 | Cost: 0.3659300 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    47 | Cost: 0.3453167 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    48 | Cost: 0.3279939 | Acc train: 0.9733333 | Acc validation: 1.0000000 \n",
            "Iter:    49 | Cost: 0.3138237 | Acc train: 0.9733333 | Acc validation: 1.0000000 \n",
            "Iter:    50 | Cost: 0.2969645 | Acc train: 0.9733333 | Acc validation: 1.0000000 \n",
            "Iter:    51 | Cost: 0.2802760 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    52 | Cost: 0.2621593 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "Iter:    53 | Cost: 0.2495680 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    54 | Cost: 0.2409833 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    55 | Cost: 0.2357028 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    56 | Cost: 0.2356819 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    57 | Cost: 0.2412892 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    58 | Cost: 0.2493107 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    59 | Cost: 0.2500630 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    60 | Cost: 0.2444099 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:     1 | Cost: 1.4609002 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     2 | Cost: 1.3367354 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     3 | Cost: 1.2024062 | Acc train: 0.4800000 | Acc validation: 0.5600000 \n",
            "Iter:     4 | Cost: 1.0523341 | Acc train: 0.4666667 | Acc validation: 0.5600000 \n",
            "Iter:     5 | Cost: 0.9222603 | Acc train: 0.5466667 | Acc validation: 0.6000000 \n",
            "Iter:     6 | Cost: 0.8481528 | Acc train: 0.7333333 | Acc validation: 0.7600000 \n",
            "Iter:     7 | Cost: 0.8377012 | Acc train: 0.7200000 | Acc validation: 0.6800000 \n",
            "Iter:     8 | Cost: 0.8992895 | Acc train: 0.5600000 | Acc validation: 0.4800000 \n",
            "Iter:     9 | Cost: 0.9278118 | Acc train: 0.5200000 | Acc validation: 0.4400000 \n",
            "Iter:    10 | Cost: 0.9258503 | Acc train: 0.5200000 | Acc validation: 0.4400000 \n",
            "Iter:    11 | Cost: 0.8819420 | Acc train: 0.5200000 | Acc validation: 0.4400000 \n",
            "Iter:    12 | Cost: 0.7922588 | Acc train: 0.6133333 | Acc validation: 0.6400000 \n",
            "Iter:    13 | Cost: 0.7383893 | Acc train: 0.8266667 | Acc validation: 0.7200000 \n",
            "Iter:    14 | Cost: 0.7267302 | Acc train: 0.8800000 | Acc validation: 0.9600000 \n",
            "Iter:    15 | Cost: 0.7268879 | Acc train: 0.8800000 | Acc validation: 0.9600000 \n",
            "Iter:    16 | Cost: 0.7330967 | Acc train: 0.8133333 | Acc validation: 0.9200000 \n",
            "Iter:    17 | Cost: 0.7355535 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    18 | Cost: 0.7361355 | Acc train: 0.8800000 | Acc validation: 0.9600000 \n",
            "Iter:    19 | Cost: 0.7394153 | Acc train: 0.8800000 | Acc validation: 0.9600000 \n",
            "Iter:    20 | Cost: 0.7428820 | Acc train: 0.9066667 | Acc validation: 0.9200000 \n",
            "Iter:    21 | Cost: 0.7537040 | Acc train: 0.7466667 | Acc validation: 0.6800000 \n",
            "Iter:    22 | Cost: 0.7474334 | Acc train: 0.8133333 | Acc validation: 0.7600000 \n",
            "Iter:    23 | Cost: 0.7418057 | Acc train: 0.8266667 | Acc validation: 0.8000000 \n",
            "Iter:    24 | Cost: 0.7363674 | Acc train: 0.8000000 | Acc validation: 0.7600000 \n",
            "Iter:    25 | Cost: 0.7398019 | Acc train: 0.6266667 | Acc validation: 0.5600000 \n",
            "Iter:    26 | Cost: 0.7320486 | Acc train: 0.6400000 | Acc validation: 0.6000000 \n",
            "Iter:    27 | Cost: 0.7132678 | Acc train: 0.8000000 | Acc validation: 0.8000000 \n",
            "Iter:    28 | Cost: 0.7141251 | Acc train: 0.7200000 | Acc validation: 0.7200000 \n",
            "Iter:    29 | Cost: 0.7410592 | Acc train: 0.5733333 | Acc validation: 0.5200000 \n",
            "Iter:    30 | Cost: 0.7839272 | Acc train: 0.5600000 | Acc validation: 0.4800000 \n",
            "Iter:    31 | Cost: 0.8080893 | Acc train: 0.5600000 | Acc validation: 0.4800000 \n",
            "Iter:    32 | Cost: 0.7428286 | Acc train: 0.5733333 | Acc validation: 0.5200000 \n",
            "Iter:    33 | Cost: 0.6915738 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    34 | Cost: 0.6512403 | Acc train: 0.8933333 | Acc validation: 0.8800000 \n",
            "Iter:    35 | Cost: 0.6457218 | Acc train: 0.9200000 | Acc validation: 1.0000000 \n",
            "Iter:    36 | Cost: 0.6508104 | Acc train: 0.8800000 | Acc validation: 0.9200000 \n",
            "Iter:    37 | Cost: 0.6547769 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    38 | Cost: 0.6300638 | Acc train: 0.9466667 | Acc validation: 0.9600000 \n",
            "Iter:    39 | Cost: 0.6199920 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    40 | Cost: 0.6102807 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    41 | Cost: 0.6033432 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    42 | Cost: 0.5932407 | Acc train: 0.9600000 | Acc validation: 1.0000000 \n",
            "Iter:    43 | Cost: 0.5795636 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    44 | Cost: 0.5680311 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    45 | Cost: 0.5560370 | Acc train: 0.9200000 | Acc validation: 0.9600000 \n",
            "Iter:    46 | Cost: 0.5473445 | Acc train: 0.9066667 | Acc validation: 0.8800000 \n",
            "Iter:    47 | Cost: 0.5338944 | Acc train: 0.9066667 | Acc validation: 0.8800000 \n",
            "Iter:    48 | Cost: 0.5167769 | Acc train: 0.9066667 | Acc validation: 0.8800000 \n",
            "Iter:    49 | Cost: 0.4954831 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    50 | Cost: 0.4785821 | Acc train: 0.9733333 | Acc validation: 1.0000000 \n",
            "Iter:    51 | Cost: 0.4657231 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    52 | Cost: 0.4575520 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    53 | Cost: 0.4504953 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    54 | Cost: 0.4368025 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    55 | Cost: 0.4293366 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    56 | Cost: 0.4164206 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    57 | Cost: 0.4199601 | Acc train: 0.9733333 | Acc validation: 0.9200000 \n",
            "Iter:    58 | Cost: 0.4087313 | Acc train: 0.9733333 | Acc validation: 0.9200000 \n",
            "Iter:    59 | Cost: 0.3959845 | Acc train: 0.9733333 | Acc validation: 0.9200000 \n",
            "Iter:    60 | Cost: 0.3527603 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:     1 | Cost: 1.4595390 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     2 | Cost: 1.3487905 | Acc train: 0.4933333 | Acc validation: 0.5600000 \n",
            "Iter:     3 | Cost: 1.2027299 | Acc train: 0.4800000 | Acc validation: 0.5600000 \n",
            "Iter:     4 | Cost: 1.0541159 | Acc train: 0.4666667 | Acc validation: 0.5600000 \n",
            "Iter:     5 | Cost: 0.9327251 | Acc train: 0.5200000 | Acc validation: 0.6400000 \n",
            "Iter:     6 | Cost: 0.8725082 | Acc train: 0.6533333 | Acc validation: 0.8000000 \n",
            "Iter:     7 | Cost: 0.8482355 | Acc train: 0.7066667 | Acc validation: 0.7600000 \n",
            "Iter:     8 | Cost: 0.8366076 | Acc train: 0.7333333 | Acc validation: 0.6800000 \n",
            "Iter:     9 | Cost: 0.8148628 | Acc train: 0.7333333 | Acc validation: 0.6800000 \n",
            "Iter:    10 | Cost: 0.7931684 | Acc train: 0.7466667 | Acc validation: 0.6800000 \n",
            "Iter:    11 | Cost: 0.7554617 | Acc train: 0.8000000 | Acc validation: 0.8000000 \n",
            "Iter:    12 | Cost: 0.7308839 | Acc train: 0.8400000 | Acc validation: 0.9600000 \n",
            "Iter:    13 | Cost: 0.7237824 | Acc train: 0.8133333 | Acc validation: 0.9200000 \n",
            "Iter:    14 | Cost: 0.7328045 | Acc train: 0.7600000 | Acc validation: 0.8000000 \n",
            "Iter:    15 | Cost: 0.7636031 | Acc train: 0.6800000 | Acc validation: 0.7600000 \n",
            "Iter:    16 | Cost: 0.7889971 | Acc train: 0.6800000 | Acc validation: 0.7200000 \n",
            "Iter:    17 | Cost: 0.7925029 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    18 | Cost: 0.7988906 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    19 | Cost: 0.7879550 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    20 | Cost: 0.7750527 | Acc train: 0.6800000 | Acc validation: 0.7200000 \n",
            "Iter:    21 | Cost: 0.7636877 | Acc train: 0.7066667 | Acc validation: 0.7600000 \n",
            "Iter:    22 | Cost: 0.7522244 | Acc train: 0.7733333 | Acc validation: 0.7600000 \n",
            "Iter:    23 | Cost: 0.7273301 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    24 | Cost: 0.7226315 | Acc train: 0.8933333 | Acc validation: 1.0000000 \n",
            "Iter:    25 | Cost: 0.7219267 | Acc train: 0.8933333 | Acc validation: 0.9200000 \n",
            "Iter:    26 | Cost: 0.7377240 | Acc train: 0.6666667 | Acc validation: 0.6400000 \n",
            "Iter:    27 | Cost: 0.7507418 | Acc train: 0.5600000 | Acc validation: 0.4800000 \n",
            "Iter:    28 | Cost: 0.7761432 | Acc train: 0.5333333 | Acc validation: 0.4400000 \n",
            "Iter:    29 | Cost: 0.7692566 | Acc train: 0.5600000 | Acc validation: 0.4400000 \n",
            "Iter:    30 | Cost: 0.7536863 | Acc train: 0.5600000 | Acc validation: 0.4800000 \n",
            "Iter:    31 | Cost: 0.7199493 | Acc train: 0.6266667 | Acc validation: 0.6000000 \n",
            "Iter:    32 | Cost: 0.6747737 | Acc train: 0.9066667 | Acc validation: 0.9200000 \n",
            "Iter:    33 | Cost: 0.6604265 | Acc train: 0.9066667 | Acc validation: 0.9600000 \n",
            "Iter:    34 | Cost: 0.6645583 | Acc train: 0.8533333 | Acc validation: 0.9200000 \n",
            "Iter:    35 | Cost: 0.6744472 | Acc train: 0.7733333 | Acc validation: 0.7600000 \n",
            "Iter:    36 | Cost: 0.7011591 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    37 | Cost: 0.7153863 | Acc train: 0.6933333 | Acc validation: 0.7200000 \n",
            "Iter:    38 | Cost: 0.6726074 | Acc train: 0.7466667 | Acc validation: 0.7600000 \n",
            "Iter:    39 | Cost: 0.6270791 | Acc train: 0.8800000 | Acc validation: 0.9200000 \n",
            "Iter:    40 | Cost: 0.6121927 | Acc train: 0.9466667 | Acc validation: 1.0000000 \n",
            "Iter:    41 | Cost: 0.6195181 | Acc train: 0.8933333 | Acc validation: 0.8400000 \n",
            "Iter:    42 | Cost: 0.6646423 | Acc train: 0.7200000 | Acc validation: 0.7600000 \n",
            "Iter:    43 | Cost: 0.7127392 | Acc train: 0.5866667 | Acc validation: 0.5200000 \n",
            "Iter:    44 | Cost: 0.7210309 | Acc train: 0.5866667 | Acc validation: 0.5200000 \n",
            "Iter:    45 | Cost: 0.6759924 | Acc train: 0.6133333 | Acc validation: 0.5600000 \n",
            "Iter:    46 | Cost: 0.6279188 | Acc train: 0.7466667 | Acc validation: 0.7600000 \n",
            "Iter:    47 | Cost: 0.5747645 | Acc train: 0.8533333 | Acc validation: 0.8400000 \n",
            "Iter:    48 | Cost: 0.5299076 | Acc train: 0.9200000 | Acc validation: 0.9600000 \n",
            "Iter:    49 | Cost: 0.5139110 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    50 | Cost: 0.5321490 | Acc train: 0.9200000 | Acc validation: 0.9200000 \n",
            "Iter:    51 | Cost: 0.5521032 | Acc train: 0.8400000 | Acc validation: 0.9200000 \n",
            "Iter:    52 | Cost: 0.5629905 | Acc train: 0.8000000 | Acc validation: 0.8400000 \n",
            "Iter:    53 | Cost: 0.5678106 | Acc train: 0.7733333 | Acc validation: 0.7200000 \n",
            "Iter:    54 | Cost: 0.5769902 | Acc train: 0.7200000 | Acc validation: 0.7200000 \n",
            "Iter:    55 | Cost: 0.5482750 | Acc train: 0.7866667 | Acc validation: 0.8400000 \n",
            "Iter:    56 | Cost: 0.5257673 | Acc train: 0.8400000 | Acc validation: 0.8400000 \n",
            "Iter:    57 | Cost: 0.4683683 | Acc train: 0.9333333 | Acc validation: 0.9200000 \n",
            "Iter:    58 | Cost: 0.4235605 | Acc train: 0.9733333 | Acc validation: 0.9600000 \n",
            "Iter:    59 | Cost: 0.3939146 | Acc train: 1.0000000 | Acc validation: 1.0000000 \n",
            "Iter:    60 | Cost: 0.3822791 | Acc train: 0.9866667 | Acc validation: 1.0000000 \n",
            "4min 18s ± 5.16 s per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "We can plot the continuous output of the variational classifier for the first two dimensions of the Iris data set."
      ],
      "metadata": {
        "id": "psGjs8nhaN89"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "plt.figure()\n",
        "cm = plt.cm.RdBu\n",
        "\n",
        "# make data for decision regions\n",
        "xx, yy = np.meshgrid(np.linspace(0.0, 1.5, 20), np.linspace(0.0, 1.5, 20))\n",
        "X_grid = [np.array([x, y]) for x, y in zip(xx.flatten(), yy.flatten())]\n",
        "\n",
        "# preprocess grid points like data inputs above\n",
        "padding = 0.3 * np.ones((len(X_grid), 1))\n",
        "X_grid = np.c_[np.c_[X_grid, padding], np.zeros((len(X_grid), 1))]  # pad each input\n",
        "normalization = np.sqrt(np.sum(X_grid ** 2, -1))\n",
        "X_grid = (X_grid.T / normalization).T  # normalize each input\n",
        "features_grid = np.array(\n",
        "    [get_angles(x) for x in X_grid]\n",
        ")  # angles for state preparation are new features\n",
        "predictions_grid = [variational_classifier(weights, bias, f) for f in features_grid]\n",
        "Z = np.reshape(predictions_grid, xx.shape)\n",
        "\n",
        "# plot decision regions\n",
        "cnt = plt.contourf(\n",
        "    xx, yy, Z, levels=np.arange(-1, 1.1, 0.1), cmap=cm, alpha=0.8, extend=\"both\"\n",
        ")\n",
        "plt.contour(\n",
        "    xx, yy, Z, levels=[0.0], colors=(\"black\",), linestyles=(\"--\",), linewidths=(0.8,)\n",
        ")\n",
        "plt.colorbar(cnt, ticks=[-1, 0, 1])\n",
        "\n",
        "# plot data\n",
        "plt.scatter(\n",
        "    X_train[:, 0][Y_train == 1],\n",
        "    X_train[:, 1][Y_train == 1],\n",
        "    c=\"b\",\n",
        "    marker=\"o\",\n",
        "    edgecolors=\"k\",\n",
        "    label=\"class 1 train\",\n",
        ")\n",
        "plt.scatter(\n",
        "    X_val[:, 0][Y_val == 1],\n",
        "    X_val[:, 1][Y_val == 1],\n",
        "    c=\"b\",\n",
        "    marker=\"^\",\n",
        "    edgecolors=\"k\",\n",
        "    label=\"class 1 validation\",\n",
        ")\n",
        "plt.scatter(\n",
        "    X_train[:, 0][Y_train == -1],\n",
        "    X_train[:, 1][Y_train == -1],\n",
        "    c=\"r\",\n",
        "    marker=\"o\",\n",
        "    edgecolors=\"k\",\n",
        "    label=\"class -1 train\",\n",
        ")\n",
        "plt.scatter(\n",
        "    X_val[:, 0][Y_val == -1],\n",
        "    X_val[:, 1][Y_val == -1],\n",
        "    c=\"r\",\n",
        "    marker=\"^\",\n",
        "    edgecolors=\"k\",\n",
        "    label=\"class -1 validation\",\n",
        ")\n",
        "\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "3weFE8x7aK0X",
        "outputId": "29c69732-4bf3-4bb8-edb9-3c2f8f467431"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "ITK1rBQBaS9m"
      },
      "execution_count": 24,
      "outputs": []
    }
  ]
}