Link_attachment_DGL.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Local Attachment with DGL\n",
    "\n",
    "#### Example Notebook for creating random graph edges (or local attachment) using Deep Graph Library (DGL) for agent based modelling\n",
    "\n",
    "Import libraries and loading sample DGL graph from previous work"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "os.environ[\"DGLBACKEND\"] = \"pytorch\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch \n",
    "import pandas as pd\n",
    "import dgl\n",
    "from dgl.data import DGLDataset\n",
    "import dgl.function as fn\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done loading data from cached files.\n"
     ]
    }
   ],
   "source": [
    "ds = dgl.data.CSVDataset('./DGL_playground/toy_dataset/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "g = ds[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Graph(num_nodes=10, num_edges=20,\n",
      "      ndata_schemes={' na2': Scheme(shape=(), dtype=torch.int64), ' na1': Scheme(shape=(), dtype=torch.int64)}\n",
      "      edata_schemes={'ea2': Scheme(shape=(), dtype=torch.int64), 'ea1': Scheme(shape=(), dtype=torch.int64), 'w': Scheme(shape=(), dtype=torch.int64)})\n"
     ]
    }
   ],
   "source": [
    "print(g)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([0, 0, 0, 1, 1, 2, 2, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8]),\n",
       " tensor([1, 2, 3, 0, 2, 0, 1, 0, 4, 5, 3, 5, 3, 4, 7, 8, 6, 8, 6, 7]))"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "g.edges()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "AddEdge adds edges randomly to the graph. For more details on the method, refer [here](https://docs.dgl.ai/generated/dgl.transforms.AddEdge.html?highlight=addedge#dgl.transforms.AddEdge)\n",
    "\n",
    "AddReverse adds the reverse edges for all existing edges on the graph. For more details on the method, refer [here](https://docs.dgl.ai/generated/dgl.transforms.AddReverse.html). Note that AddReverse does not check for duplicate edges but rather adds a reverse edge for every existing edge."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dgl import AddEdge, AddReverse"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ratio defines number of new edges to add as a proportion of existing edges. \n",
    "# Here we want to add as many edges as number of nodes\n",
    "ratio = g.num_nodes()/g.num_edges()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "new_g = AddEdge(ratio=ratio)(g)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([0, 0, 0, 1, 1, 2, 2, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 3, 3, 9, 8,\n",
       "         5, 1, 3, 8, 5, 7]),\n",
       " tensor([1, 2, 3, 0, 2, 0, 1, 0, 4, 5, 3, 5, 3, 4, 7, 8, 6, 8, 6, 7, 5, 9, 1, 0,\n",
       "         9, 8, 8, 5, 8, 8]))"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_g.edges()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "new_g2 = AddReverse()(new_g)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([0, 0, 0, 1, 1, 2, 2, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 3, 3, 9, 8,\n",
       "         5, 1, 3, 8, 5, 7, 1, 2, 3, 0, 2, 0, 1, 0, 4, 5, 3, 5, 3, 4, 7, 8, 6, 8,\n",
       "         6, 7, 5, 9, 1, 0, 9, 8, 8, 5, 8, 8]),\n",
       " tensor([1, 2, 3, 0, 2, 0, 1, 0, 4, 5, 3, 5, 3, 4, 7, 8, 6, 8, 6, 7, 5, 9, 1, 0,\n",
       "         9, 8, 8, 5, 8, 8, 0, 0, 0, 1, 1, 2, 2, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7,\n",
       "         8, 8, 3, 3, 9, 8, 5, 1, 3, 8, 5, 7]))"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_g2.edges()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "to_simple function removes duplicate edges from the graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 3, 3, 4, 4, 5, 5, 5, 5, 6, 6, 7,\n",
       "         7, 8, 8, 8, 8, 8, 8, 9, 9, 9]),\n",
       " tensor([1, 2, 3, 8, 0, 2, 8, 9, 0, 1, 0, 4, 5, 8, 9, 3, 5, 3, 4, 8, 9, 7, 8, 6,\n",
       "         8, 0, 1, 3, 5, 6, 7, 1, 3, 5]))"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_g3 = dgl.to_simple(new_g2, return_counts='cnt')\n",
    "new_g3.edges()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "ptm",
   "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": "3.11.4"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}