clustered_federated_learning.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "from IPython.display import clear_output\n",
    "import os\n",
    "from copy import deepcopy\n",
    "\n",
    "import torch\n",
    "from torch.utils.data import DataLoader\n",
    "from torchvision import datasets, transforms\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from models import ConvNet\n",
    "from helper import ExperimentLogger, display_train_stats\n",
    "from fl_devices import Server, Client\n",
    "from data_utils import split_noniid, CustomSubset\n",
    "\n",
    "\n",
    "torch.manual_seed(42)\n",
    "np.random.seed(42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will run experiments with the EMNIST data set of handwritten characters. We simulate a scenario with 10 clients and distribute the data among the clients in a non-iid way accoding to a dirichlet distribution parametrized with alpha=1.0 . The distribution of data is shown below."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/sattler/anaconda3/lib/python3.7/site-packages/torchvision/datasets/mnist.py:45: UserWarning: train_labels has been renamed targets\n",
      "  warnings.warn(\"train_labels has been renamed targets\")\n"
     ]
    }
   ],
   "source": [
    "N_CLIENTS = 10\n",
    "DIRICHLET_ALPHA = 1.0\n",
    "\n",
    "\n",
    "data = datasets.EMNIST(root=\".\", split=\"byclass\", download=True)\n",
    "\n",
    "mapp = np.array(['0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C',\n",
    "       'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P',\n",
    "       'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z', 'a', 'b', 'c',\n",
    "       'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n', 'o', 'p',\n",
    "       'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z'], dtype='<U1')\n",
    "\n",
    "idcs = np.random.permutation(len(data))\n",
    "train_idcs, test_idcs = idcs[:10000], idcs[10000:20000]\n",
    "train_labels = data.train_labels.numpy()\n",
    "\n",
    "client_idcs = split_noniid(train_idcs, train_labels, alpha=DIRICHLET_ALPHA, n_clients=N_CLIENTS)\n",
    "\n",
    "client_data = [CustomSubset(data, idcs) for idcs in client_idcs]\n",
    "test_data = CustomSubset(data, test_idcs, transforms.Compose([transforms.ToTensor()]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x216 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(20,3))\n",
    "plt.hist([train_labels[idc]for idc in client_idcs], stacked=True, \n",
    "         bins=np.arange(min(train_labels)-0.5, max(train_labels) + 1.5, 1),\n",
    "        label=[\"Client {}\".format(i) for i in range(N_CLIENTS)])\n",
    "plt.xticks(np.arange(62), mapp)\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next, we simulate a clustering structure in the client population, by rotating the data for half of the first 5 clients by 180 degree. We display 10 data samples from the 1st and the 6th client for illustration."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i, client_datum in enumerate(client_data):\n",
    "    if i<5:\n",
    "        client_datum.subset_transform = transforms.Compose([transforms.RandomRotation((180,180)),\n",
    "                                                      transforms.ToTensor()])\n",
    "    else:\n",
    "        client_datum.subset_transform = transforms.Compose([transforms.ToTensor()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "clients = [Client(ConvNet, lambda x : torch.optim.SGD(x, lr=0.1, momentum=0.9), dat, idnum=i) \n",
    "           for i, dat in enumerate(client_data)]\n",
    "server = Server(ConvNet, test_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Client 0:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x72 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Client 5:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x72 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for client in [clients[0], clients[5]]:\n",
    "    x, y = iter(client.train_loader).next()\n",
    "\n",
    "    print(\"Client {}:\".format(client.id))\n",
    "    plt.figure(figsize=(15,1))\n",
    "    for i in range(10):\n",
    "        plt.subplot(1,10,i+1)\n",
    "        plt.imshow(x[i,0].numpy().T, cmap=\"Greys\")\n",
    "        plt.title(\"Label: {}\".format(mapp[y[i].item()]))\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now everything is set up to run our Clustered Federated Learning algorithm. During training, we will track the mean and std client accuracies, as well as the average and maximum client update norms."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "COMMUNICATION_ROUNDS = 80\n",
    "EPS_1 = 0.4\n",
    "EPS_2 = 1.6\n",
    "    \n",
    "    \n",
    "cfl_stats = ExperimentLogger()\n",
    "    \n",
    "cluster_indices = [np.arange(len(clients)).astype(\"int\")]\n",
    "client_clusters = [[clients[i] for i in idcs] for idcs in cluster_indices]\n",
    "\n",
    "\n",
    "for c_round in range(1, COMMUNICATION_ROUNDS+1):\n",
    "\n",
    "    if c_round == 1:\n",
    "        for client in clients:\n",
    "            client.synchronize_with_server(server)\n",
    "            \n",
    "    participating_clients = server.select_clients(clients, frac=1.0)\n",
    "\n",
    "    for client in participating_clients:\n",
    "        train_stats = client.compute_weight_update(epochs=1)\n",
    "        client.reset()\n",
    "\n",
    "    similarities = server.compute_pairwise_similarities(clients)\n",
    "\n",
    "    cluster_indices_new = []\n",
    "    for idc in cluster_indices:\n",
    "        max_norm = server.compute_max_update_norm([clients[i] for i in idc])\n",
    "        mean_norm = server.compute_mean_update_norm([clients[i] for i in idc])\n",
    "             \n",
    "        if mean_norm<EPS_1 and max_norm>EPS_2 and len(idc)>2 and c_round>20:\n",
    "            \n",
    "            server.cache_model(idc, clients[idc[0]].W, acc_clients)\n",
    "            \n",
    "            c1, c2 = server.cluster_clients(similarities[idc][:,idc]) \n",
    "            cluster_indices_new += [c1, c2]\n",
    "             \n",
    "            cfl_stats.log({\"split\" : c_round})\n",
    "\n",
    "        else:\n",
    "            cluster_indices_new += [idc]\n",
    "        \n",
    "        \n",
    "    cluster_indices = cluster_indices_new\n",
    "    client_clusters = [[clients[i] for i in idcs] for idcs in cluster_indices]\n",
    "\n",
    "    server.aggregate_clusterwise(client_clusters)\n",
    "\n",
    "    acc_clients = [client.evaluate() for client in clients]\n",
    "    \n",
    "    cfl_stats.log({\"acc_clients\" : acc_clients, \"mean_norm\" : mean_norm, \"max_norm\" : max_norm,\n",
    "                  \"rounds\" : c_round, \"clusters\" : cluster_indices})\n",
    "    \n",
    "    \n",
    "    display_train_stats(cfl_stats, EPS_1, EPS_2, COMMUNICATION_ROUNDS)\n",
    "\n",
    "    \n",
    "for idc in cluster_indices:    \n",
    "    server.cache_model(idc, clients[idc[0]].W, acc_clients)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The training process resulted in multiple models for every client: A Federated Learning base model as well as more specialized models for the different clusters.  We can now compare their accuracies on the clients' validation sets, and assign each client the model which performed best."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "results = np.zeros([10, len(server.model_cache)])\n",
    "for i, (idcs, W, accs) in enumerate(server.model_cache):\n",
    "    results[idcs, i] = np.array(accs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style  type=\"text/css\" >\n",
       "    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col0 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col1 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col2 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col3 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col4 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col5 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col6 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col7 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col8 {\n",
       "            background-color:  yellow;\n",
       "        }    #T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col9 {\n",
       "            background-color:  yellow;\n",
       "        }</style><table id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4\" ><thead>    <tr>        <th class=\"blank level0\" ></th>        <th class=\"col_heading level0 col0\" >Client 0</th>        <th class=\"col_heading level0 col1\" >Client 1</th>        <th class=\"col_heading level0 col2\" >Client 2</th>        <th class=\"col_heading level0 col3\" >Client 3</th>        <th class=\"col_heading level0 col4\" >Client 4</th>        <th class=\"col_heading level0 col5\" >Client 5</th>        <th class=\"col_heading level0 col6\" >Client 6</th>        <th class=\"col_heading level0 col7\" >Client 7</th>        <th class=\"col_heading level0 col8\" >Client 8</th>        <th class=\"col_heading level0 col9\" >Client 9</th>    </tr></thead><tbody>\n",
       "                <tr>\n",
       "                        <th id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4level0_row0\" class=\"row_heading level0 row0\" >FL Model</th>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col0\" class=\"data row0 col0\" >0.655340</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col1\" class=\"data row0 col1\" >0.692884</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col2\" class=\"data row0 col2\" >0.683333</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col3\" class=\"data row0 col3\" >0.675105</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col4\" class=\"data row0 col4\" >0.503759</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col5\" class=\"data row0 col5\" >0.600000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col6\" class=\"data row0 col6\" >0.508671</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col7\" class=\"data row0 col7\" >0.683616</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col8\" class=\"data row0 col8\" >0.575472</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row0_col9\" class=\"data row0 col9\" >0.638298</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4level0_row1\" class=\"row_heading level0 row1\" >Model 0</th>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col0\" class=\"data row1 col0\" >0.757282</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col1\" class=\"data row1 col1\" >0.782772</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col2\" class=\"data row1 col2\" >0.787500</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col3\" class=\"data row1 col3\" >0.763713</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col4\" class=\"data row1 col4\" >0.654135</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col5\" class=\"data row1 col5\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col6\" class=\"data row1 col6\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col7\" class=\"data row1 col7\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col8\" class=\"data row1 col8\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row1_col9\" class=\"data row1 col9\" >0.000000</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4level0_row2\" class=\"row_heading level0 row2\" >Model 1</th>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col0\" class=\"data row2 col0\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col1\" class=\"data row2 col1\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col2\" class=\"data row2 col2\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col3\" class=\"data row2 col3\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col4\" class=\"data row2 col4\" >0.000000</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col5\" class=\"data row2 col5\" >0.694118</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col6\" class=\"data row2 col6\" >0.595376</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col7\" class=\"data row2 col7\" >0.745763</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col8\" class=\"data row2 col8\" >0.740566</td>\n",
       "                        <td id=\"T_82f886ea_abb2_11ea_a41a_378b73607ee4row2_col9\" class=\"data row2 col9\" >0.744681</td>\n",
       "            </tr>\n",
       "    </tbody></table>"
      ],
      "text/plain": [
       "<pandas.io.formats.style.Styler at 0x7fd7fa265390>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "frame = pd.DataFrame(results, columns=[\"FL Model\"]+[\"Model {}\".format(i) \n",
    "                                                    for i in range(results.shape[1]-1)],\n",
    "            index = [\"Client {}\".format(i) for i in range(results.shape[0])])\n",
    "\n",
    "def highlight_max(s):\n",
    "    is_max = s == s.max()\n",
    "    return ['background-color: yellow' if v else '' for v in is_max]\n",
    "frame.T.style.apply(highlight_max)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we can see, clustering improoved the accuracy for all clients by about 10%."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}