torch_erf.ipynb

torch_erf.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "torch_erf.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyO7ZPv+XfCDUHidfyQ3b9au",
      "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/gist/tomonari-masada/913d6a910c13c3a9f733364c7be124ff/torch_erf.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import torch\n",
        "from torch.special import erf\n",
        "\n",
        "# 微分係数を計算する位置\n",
        "a = 1.0\n",
        "\n",
        "# erfの自動微分\n",
        "x = torch.tensor([a], requires_grad=True)\n",
        "y = erf(x)\n",
        "y.backward()\n",
        "print(x.grad.item())\n",
        "\n",
        "# 答え合わせ（erfを式の計算で微分しておいたものにaを当てはめる。）\n",
        "print(2 / np.sqrt(np.pi) * np.exp(- a ** 2))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "X24lm_yqPJvd",
        "outputId": "6197b06a-38e5-4612-a3e3-f88859402d88"
      },
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0.41510751843452454\n",
            "0.4151074974205947\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "hu5XKjrGez0A"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}