{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Usage Example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np \n",
    "import pandas as pd \n",
    "import torch\n",
    "from anglepy import ANGLE\n",
    "from anglepy.metrics import mean_absolute_angular_deviation, circular_mean_directional_error, mean_circular_crps, accuracy, med_err\n",
    "from anglepy.kernels import c2_wendland\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tqdm.auto import tqdm\n",
    "import math"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Indian Wind Dataset - Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "trusted": true
   },
   "outputs": [],
   "source": [
    "df = pd.read_csv('India_Wind_Data.csv')\n",
    "random_state=18\n",
    "device= torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "X, y = df[['Latitude', 'Longitude']], df['Direction_Rad']\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, shuffle=True, random_state=random_state)\n",
    "X_train, X_test, y_train, y_test = torch.from_numpy(X_train.values).float().to(device), torch.from_numpy(X_test.values).float().to(device), torch.from_numpy(y_train.values).float().to(device), torch.from_numpy(y_test.values).float().to(device)\n",
    "y_train = y_train.unsqueeze(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "trusted": true
   },
   "outputs": [],
   "source": [
    "# ANGLE-Geodesic\n",
    "\n",
    "accuracies, mederrs, crpss, cmdes, maads, times = [], [], [], [], [], []\n",
    "\n",
    "for _ in tqdm(range(50)):\n",
    "    model = ANGLE(\n",
    "        X_train, y_train, \n",
    "        num_layer=3, hidden_dim=100, noise_dim=64, \n",
    "            add_bn=True, resblock=False, noise_std=1,\n",
    "            print_every_nepoch=10000, print_times_per_epoch=0, \n",
    "            lr=0.05, num_epochs=500, batch_size=None, \n",
    "            standardize=False, device=device,\n",
    "            dist_method='geodesic', verbose=False, circular_indices=None,\n",
    "            kernel_func=c2_wendland(), circular_projection='atan2'\n",
    "    )\n",
    "    \n",
    "    y_pred = model.predict(X_test)\n",
    "    y_pred_ensemble = model.sample(X_test, sample_size=100)\n",
    "    maad_mean = mean_absolute_angular_deviation(y_pred, y_test, return_degrees=True).item()\n",
    "    cmde = circular_mean_directional_error(y_pred, y_test).item()\n",
    "    crps = mean_circular_crps(y_pred_ensemble, y_test, return_degrees=True)[0]\n",
    "    acc = accuracy(y_pred, y_test, theta=torch.pi/6)\n",
    "    median_err = med_err(y_pred, y_test)\n",
    "    accuracies.append(round(acc,3))\n",
    "    mederrs.append(round(median_err,3))\n",
    "    crpss.append(round(crps, 3))\n",
    "    maads.append(round(maad_mean, 3))\n",
    "    cmdes.append(round(cmde,3))\n",
    "print(f\"{np.mean(accuracies):.2f}\\pm{np.std(accuracies):.2f}\")\n",
    "print(f\"{np.mean(mederrs):.2f}\\pm{np.std(mederrs):.2f}\")\n",
    "print(f\"{np.mean(crpss):.2f}\\pm{np.std(crpss):.2f}\")\n",
    "print(f\"{np.mean(maads):.2f}\\pm{np.std(maads):.2f}\")\n",
    "print(f\"{np.mean(cmdes):.2f}\\pm{np.std(cmdes):.2f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "trusted": true
   },
   "outputs": [],
   "source": [
    "# ANGLE-Chordal\n",
    "accuracies, mederrs, crpss, cmdes, maads, times = [], [], [], [], [], []\n",
    "\n",
    "for _ in tqdm(range(50)):\n",
    "    model = ANGLE(\n",
    "        X_train, y_train, \n",
    "        num_layer=3, hidden_dim=100, noise_dim=64, \n",
    "            add_bn=True, resblock=False, gamma=1, noise_std=1,\n",
    "            print_every_nepoch=10000, print_times_per_epoch=0, \n",
    "            lr=0.05, num_epochs=800, batch_size=None, \n",
    "            standardize=False, device=device,\n",
    "            dist_method='chordal', verbose=False, circular_indices=None\n",
    "    )\n",
    "    \n",
    "    y_pred = model.predict(X_test)\n",
    "    y_pred_ensemble = model.sample(X_test, sample_size=100)\n",
    "    maad_mean = mean_absolute_angular_deviation(y_pred, y_test, return_degrees=True).item()\n",
    "    cmde = circular_mean_directional_error(y_pred, y_test).item()\n",
    "    crps = mean_circular_crps(y_pred_ensemble, y_test, return_degrees=True)[0]\n",
    "    acc = accuracy(y_pred, y_test, theta=torch.pi/6)\n",
    "    median_err = med_err(y_pred, y_test)\n",
    "    accuracies.append(round(acc,3))\n",
    "    mederrs.append(round(median_err,3))\n",
    "    crpss.append(round(crps, 3))\n",
    "    maads.append(round(maad_mean, 3))\n",
    "    cmdes.append(round(cmde,3))\n",
    "\n",
    "print(f\"{np.mean(accuracies):.2f}\\pm{np.std(accuracies):.2f}\")\n",
    "print(f\"{np.mean(mederrs):.2f}\\pm{np.std(mederrs):.2f}\")\n",
    "print(f\"{np.mean(crpss):.2f}\\pm{np.std(crpss):.2f}\")\n",
    "print(f\"{np.mean(maads):.2f}\\pm{np.std(maads):.2f}\")\n",
    "print(f\"{np.mean(cmdes):.2f}\\pm{np.std(cmdes):.2f}\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Indian Wind Dataset - Sufficient Dimension Reduction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('India_Wind_Data.csv')\n",
    "random_state=18\n",
    "device= torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "X, y = df[['Latitude', 'Longitude']], df['Direction_Rad']\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, shuffle=True, random_state=random_state)\n",
    "X_train, X_test, y_train, y_test = torch.from_numpy(X_train.values).float().to(device), torch.from_numpy(X_test.values).float().to(device), torch.from_numpy(y_train.values).float().to(device), torch.from_numpy(y_test.values).float().to(device)\n",
    "y_train = y_train.unsqueeze(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = ANGLE(\n",
    "        X_train, y_train, \n",
    "        num_layer=3, hidden_dim=100, noise_dim=64, \n",
    "            add_bn=True, resblock=False, gamma=1, noise_std=1,\n",
    "            print_every_nepoch=10000, print_times_per_epoch=0, \n",
    "            lr=0.05, num_epochs=800, batch_size=None, \n",
    "            standardize=True, device=device,\n",
    "            dist_method='chordal', verbose=False, circular_indices=None, sdr=True, reduced_dim=1\n",
    "    )\n",
    "beta_matrix = model.beta_proj\n",
    "print(beta_matrix.shape) \n",
    "# Output: (1, d)\n",
    "\n",
    "# Project your data to 1D for fitting stage-2 ANGLE, or plotting\n",
    "reduced_X = X_train @ beta_matrix.T"
   ]
  }
 ],
 "metadata": {
  "kaggle": {
   "accelerator": "none",
   "dataSources": [],
   "dockerImageVersionId": 28755,
   "isGpuEnabled": false,
   "isInternetEnabled": false,
   "language": "python",
   "sourceType": "notebook"
  },
  "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": "3.12.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}