{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "internals": {
     "slide_helper": "subslide_end",
     "slide_type": "subslide"
    },
    "slide_helper": "slide_end",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Aprendiendo Machine Learning con el Mundial\n",
    "===========================================\n",
    "\n",
    "Nombre alternativo de la charla: Antonio\n",
    "\n",
    "Juan Pedro Fisanotti"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "internals": {
     "slide_type": "subslide"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Objetivo de la charla:\n",
    "======================"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "internals": {
     "frag_number": 2
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- Aprender algunas nociones básicas de Machine Learning\n",
    "- Reforzar esos conocimientos con un ejemplo real"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Machine Learning\n",
    "================"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Qué es?\n",
    "======="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- Rama de la inteligencia artificial\n",
    "- Programas que \"aprenden desde los datos\"\n",
    "- Cercana a la estadística y métodos matemáticos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Aprendizaje no supervisado\n",
    "=========================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Idea principal: reconocer patrones o **agrupaciones** en los datos.\n",
    "\n",
    "No interpreta los grupos, eso nos toca a nosotros."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/unsupervised_learning_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/unsupervised_learning_2.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Aprendizaje supervisado\n",
    "======================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Idea principal: encontrar una **función** que explique los datos de salida a partir de los datos de entrada.\n",
    "\n",
    "Nosotros tenemos **ambos** grupos de datos."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/supervised_learning_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Aprendizaje supervisado - Regresión\n",
    "==================================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Una función corriente y moliente."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/regression_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/regression_2.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Aprendizaje supervisado - Clasificación\n",
    "======================================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Una función que devuelva **etiquetas**."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/clasification_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/clasification_2.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Algunas nociones importantes de aprendizaje supervisado\n",
    "======================================================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Idea de hipótesis\n",
    "================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/idea_hipotesis_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/idea_hipotesis_2.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Es una función **candidata** a ser la que buscamos para predecir las salidas. \n",
    "\n",
    "Los algoritmos van a probar **muchas** hipótesis, y devolvernos la mejor que encuentren."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Idea de modelo\n",
    "=============="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/idea_modelo_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/idea_modelo_2.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Es el tipo de función, el **molde**, que vamos a utilizar para armar nuestras hipótesis.\n",
    "\n",
    "Por lo general, un algoritmo específico de IA sabe trabajar con un tipo específico de funciones, o grupo de tipos similares."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Entrenamiento y predicción\n",
    "====================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/entrenamiento.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/prediccion.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "* Primero entrenamos para encontrar una buena hipótesis\n",
    "* Luego podemos predecir nuevos casos usando la hipótesis encontrada"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Modelos paramétricos y no paramétricos\n",
    "======================================"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Hablando mal y pronto:\n",
    "\n",
    "* **Modelo paramétrico**: arma una f() con algunos parámetros, y eso **solo** basta para hacer predicciones.\n",
    "\n",
    "* **Modelo no paramétrico**: tiene una f() que consulta los datos. Para predecir hace falta la f() **más** los datos."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/parametricos_vs_no_parametricos.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Sobreentrenamiento\n",
    "=================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/high_variance_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/high_variance_2.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/high_variance_3.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/high_variance_4.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/high_variance_5.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/high_variance_6.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Algunos algoritmos de aprendizaje supervisado\n",
    "=============================================\n",
    "\n",
    "(ejemplos, hay muchos)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "K-Vecinos\n",
    "========="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/k_neighbors_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/k_neighbors_2.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/k_neighbors_3.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/k_neighbors_4.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "- Es **no paramétrico**. Necesita los datos siempre.\n",
    "- Es muy **simple**, no tiene tiempo de entrenamiento."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Redes neuronales\n",
    "================"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/neural_networks_1.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "![](files/neural_networks_2.svg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "- Es **paramétrico**, aprende una función y con eso puede predecir, sin llevarse los datos.\n",
    "- Hay que entrenarlas, y es un proceso **pesado** y **lento**.\n",
    "- Pueden aprender funciones super complejas (técnicamente, **cualquier** función)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "El Mundial\n",
    "=========="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- Gente que patea una **pelota**.\n",
    "- Aparentemente, meter la pelota en el **arco** del otro equipo, es bueno. No se sabe bien por qué.\n",
    "- Cada tanto hay una competencia para ver qué país tiene los mejores peloteadores."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "El problema\n",
    "==========="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Cometí el error de hablar demasiado en un almuerzo con la familia de mi novia (me suele pasar).\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- Prode familiar!!\n",
    "- Participación obligatoria\n",
    "- El perdedor cocina para todos\n",
    "- Se tanto de cocina como de fútbol"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "La solución\n",
    "==========="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- Armar un **clasificador**\n",
    "- Entradas: partidos\n",
    "- Salidas: quiénes ganan los partidos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Paso 1: obtener datos\n",
    "====================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- Suele renegarse. Renegué (lugar, formato, completitud, etc).\n",
    "- Mejor fuente: Wikipedia\n",
    "- Formato: html, no muy estandarizado --> scrapeo + copy/paste --> csv"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Paso 2: pre-procesar los datos\n",
    "=============================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- \"**Numerizar**\" nombres de equipos --> estadísticas\n",
    "- **Normalizar** valores\n",
    "- Contrarrestar bias de **orden** de equipo\n",
    "- **Foco** en lo que pueda predecir mejor. Ganados vs empatados."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Paso 3: los datos son clasificables?\n",
    "===================================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- No se pueden graficar todas las dimensiones\n",
    "- Confirmadas algunas intuiciones obvias"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Paso 4: elegir algoritmo\n",
    "========================"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- K-vecinos?\n",
    "- Redes neuronales? "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Paso 5: entrenamiento\n",
    "====================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- Back propagation\n",
    "- Evitar sobreentrenamiento"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Paso 6: predecir!\n",
    "================="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- No olvidar re-pre-procesar, normalizar, etc.\n",
    "- Ida y vuelta\n",
    "- Todos contra todos\n",
    "- Re-entrenamiento entre etapas"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Ya que estamos, competir online en El Ega."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "El resultado\n",
    "============"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Prode familiar:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "**Primer puesto!!**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "El Ega (+200 personas):"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "**Primer puesto!!!!!!!**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Conclusiones\n",
    "============"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Machine Learning es algo que puede **fácilmente** aplicarse a problemas reales y no tan grandes, con resultados **suficientemente** buenos (mejores que un humano) para el problema.\n",
    "\n",
    "Hay que perderle el miedo."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "?\n",
    "===="
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true,
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "Links útiles\n",
    "============\n",
    "\n",
    "El código del predictor del mundial visualizado con ejemplos de datos y gráficos:\n",
    "\n",
    "http://nbviewer.ipython.org/github/fisadev/world_cup_learning/blob/master/learn.ipynb\n",
    "\n",
    "El código fuente de eso mismo en github: \n",
    "\n",
    "https://github.com/fisadev/world_cup_learning"
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2+"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}