01_qm9_final_postprocessing2_pca_analysis_train_test_split.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "c7c0172d-ad8e-42f8-b796-b4ec7833e33e",
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%config Completer.use_jedi = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "f387788b-c16a-45d2-8e70-e309ca3b944d",
   "metadata": {},
   "outputs": [],
   "source": [
    "from datetime import datetime\n",
    "from collections import Counter\n",
    "from copy import deepcopy\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from scipy.stats import spearmanr\n",
    "from pathlib import Path\n",
    "import pickle\n",
    "import sys\n",
    "\n",
    "import numpy as np\n",
    "from monty.json import MSONable\n",
    "from pymatgen.core.structure import Molecule"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "ab130d4f-c85e-4657-8b5f-5495d56be4d9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# https://gist.github.com/x94carbone/f5201b1c44963ff9453b9cc1d5f768ac\n",
    "sys.path.append(str(Path.home() / Path(\"local\")))\n",
    "from mpl_utils import MPLAdjutant\n",
    "adj = MPLAdjutant()\n",
    "adj.set_defaults()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "e919c247-657a-4d05-8f0f-56ecd45b1e3d",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "\n",
    "def save_json(d, path):\n",
    "    with open(path, 'w') as outfile:\n",
    "        json.dump(d, outfile, indent=4, sort_keys=True)\n",
    "\n",
    "def read_json(path):\n",
    "    with open(path, 'r') as infile:\n",
    "        dat = json.load(infile)\n",
    "    return dat"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1ecad749-6306-40bf-bb70-d9e5f7378648",
   "metadata": {},
   "source": [
    "Append the `home` path of this project."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "77708ce8-462d-4d5a-9f97-6ae089a9a852",
   "metadata": {},
   "outputs": [],
   "source": [
    "sys.path.append(str(Path.cwd().parent))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4a8d5d44-b24b-45c3-b82e-a6542ee345f4",
   "metadata": {},
   "source": [
    "# PCA"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9634bebc-8513-47eb-b231-c1d27a8f685d",
   "metadata": {},
   "source": [
    "Load in the previously processed `pickle` data. These contain the features (ACSF) and targets (spectra) on a common grid for each of the three absorbing atom types of interest: C, N and O."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "94bcf6d6-f9b5-4f78-af8f-8f85b21ac80e",
   "metadata": {},
   "outputs": [],
   "source": [
    "O_data = pickle.load(open(\"data/qm9/XANES-220622-ACSF-O.pkl\", \"rb\"))\n",
    "N_data = pickle.load(open(\"data/qm9/XANES-220622-ACSF-N.pkl\", \"rb\"))\n",
    "C_data = pickle.load(open(\"data/qm9/XANES-220622-ACSF-C.pkl\", \"rb\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "820d2e6b-4513-4bbf-a70e-3a4234797d84",
   "metadata": {},
   "source": [
    "Print some diagonstics..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "5d57ef9b-85a7-41a2-9d69-d6bd77eaa22d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "O database ACSF & spectrum shapes: (186813, 155) & (186813, 200)\n",
      "N database ACSF & spectrum shapes: (139058, 155) & (139058, 200)\n",
      "C database ACSF & spectrum shapes: (841912, 155) & (841912, 200)\n"
     ]
    }
   ],
   "source": [
    "print(f\"O database ACSF & spectrum shapes: {O_data['x'].shape} & {O_data['y'].shape}\")\n",
    "print(f\"N database ACSF & spectrum shapes: {N_data['x'].shape} & {N_data['y'].shape}\")\n",
    "print(f\"C database ACSF & spectrum shapes: {C_data['x'].shape} & {C_data['y'].shape}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6e37598f-b54d-47ba-8091-5c88895aa3a5",
   "metadata": {},
   "source": [
    "Put everything in a dictionary for simplicity."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "3ac006f3-374d-4ace-9a6f-a1dda5da0bbe",
   "metadata": {},
   "outputs": [],
   "source": [
    "data = {\"O\": O_data, \"N\": N_data, \"C\": C_data}"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1fbc6c31-f03b-4e3c-960b-1495ee04747f",
   "metadata": {},
   "source": [
    "Load in the FEFF failure example."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "51acfac9-8cdd-46ab-b918-69bc747258bf",
   "metadata": {},
   "source": [
    "## Execute and plot"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ad45c503-d1c9-474d-a43c-346fec3c3653",
   "metadata": {},
   "source": [
    "We perform basic analysis and Principal Component Analysis to validate that our dataset look sensible."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5aceb424-eb8b-4a82-a5e7-2848079c81c1",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.decomposition import PCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9c59c143-e615-4f82-bd05-aa345b1d5660",
   "metadata": {},
   "outputs": [],
   "source": [
    "pca_results = dict()\n",
    "for key, dat in data.items():\n",
    "    pca_acsf = PCA(2)\n",
    "    w_acsf = pca_acsf.fit_transform(dat[\"x\"])\n",
    "    pca_spectra = PCA(2)\n",
    "    w_spectra = pca_spectra.fit_transform(dat[\"y\"])\n",
    "    pca_results[key] = {\n",
    "        \"pca_ACSF\": pca_acsf,\n",
    "        \"pca_spectra\": pca_spectra,\n",
    "        \"w_ACSF\": w_acsf,\n",
    "        \"w_spectra\": w_spectra\n",
    "    }"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1b9221bf-e60d-415a-816a-9bccfd54100b",
   "metadata": {},
   "source": [
    "We make a production-quality plot of the data here:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "004805be-8f07-4aea-ae44-015fbb52e971",
   "metadata": {},
   "outputs": [],
   "source": [
    "fig, axs = plt.subplots(1, 3, figsize=(6, 2))\n",
    "\n",
    "N = None\n",
    "key_ordering = [\"C\", \"N\", \"O\"]\n",
    "\n",
    "for key, ax in zip(key_ordering, axs):\n",
    "    w_acsf = pca_results[key][\"w_ACSF\"]\n",
    "    w_spectra = pca_results[key][\"w_spectra\"]\n",
    "    ax.scatter(w_acsf[:N, 0], w_acsf[:N, 1], alpha=0.02, c=w_spectra[:N, 0], s=1, cmap=\"rainbow\", rasterized=True)\n",
    "    ax.spines[\"right\"].set_visible(False)\n",
    "    ax.spines[\"top\"].set_visible(False)\n",
    "    ax.spines[\"left\"].set_linewidth(0.5)\n",
    "    ax.spines[\"bottom\"].set_linewidth(0.5)\n",
    "    ax.tick_params(left=False, bottom=False)\n",
    "    ax.set_yticklabels([])\n",
    "    ax.set_xticklabels([])\n",
    "    ax.set_title(key)\n",
    "\n",
    "axs[0].set_ylabel(r\"ACSF $z_2$\")\n",
    "axs[1].set_xlabel(r\"ACSF $z_1$\")\n",
    "    \n",
    "# plt.savefig(\"Figures/qm9_pca.svg\", bbox_inches=\"tight\", dpi=300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "05898365-055c-4638-8aae-4dc850c43e2f",
   "metadata": {},
   "source": [
    "# Train/validation-test split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "7b048bcb-953a-44d1-9081-d7c728cb9e39",
   "metadata": {},
   "outputs": [],
   "source": [
    "from xas_nne.qm9 import remove_zwitter_ions_, random_split, split_qm9_data_by_number_of_total_atoms, split_qm9_data_by_number_of_absorbers, atom_count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4f387ab2-824e-4234-be49-100d6ce2195e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def pca_decomposition_inputs(data, max_pca=50, capture_variance=0.999):\n",
    "    data = deepcopy(data)\n",
    "    pca = PCA(max_pca)\n",
    "    pca.fit(data[\"train\"][\"x\"])\n",
    "    max_column = np.where(np.cumsum(pca.explained_variance_ratio_) > capture_variance)[0][0]\n",
    "    print(f\"maxcol={max_column}\")\n",
    "    for key in [\"train\", \"val\", \"test\"]:\n",
    "        data[key][\"x\"] = pca.transform(data[key][\"x\"])[:, :max_column+1]\n",
    "    data[\"pca\"] = pca\n",
    "    return data, max_column"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "93a986f6-3812-4c14-a33c-bf79d633bd14",
   "metadata": {},
   "source": [
    "## Remove Zwitter ions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c37d795a-cfef-4220-85bb-7b004ca00946",
   "metadata": {},
   "outputs": [],
   "source": [
    "for CENTRAL_ATOM in [\"C\", \"N\", \"O\"]:\n",
    "    print(f\"{CENTRAL_ATOM} original length {len(data[CENTRAL_ATOM]['x'])}\")\n",
    "    remove_zwitter_ions_(data[CENTRAL_ATOM])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f5d23cb0-a9f0-4961-b19a-180d9fc123ad",
   "metadata": {},
   "source": [
    "## Random split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b0989a85-45c1-47f2-b27b-598be13067fd",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "for CENTRAL_ATOM in [\"C\", \"N\", \"O\"]:\n",
    "    prop = 0.1 if CENTRAL_ATOM != \"C\" else 0.05\n",
    "    parsed_data = random_split(data[CENTRAL_ATOM], prop_test=prop, prop_val=prop)\n",
    "    now = datetime.now().strftime(\"%y%m%d\")\n",
    "    fname = f\"data/qm9/ml_ready/random_splits/XANES-{now}-ACSF-{CENTRAL_ATOM}-RANDOM-SPLITS.pkl\"\n",
    "    pickle.dump(parsed_data, open(fname, \"wb\"), protocol=pickle.HIGHEST_PROTOCOL)\n",
    "    \n",
    "    parsed_data, max_column = pca_decomposition_inputs(parsed_data)\n",
    "    fname = f\"data/qm9/ml_ready/random_splits/XANES-{now}-ACSF-{CENTRAL_ATOM}-RANDOM-SPLITS-PCA-decomp-maxcol-{max_column}.pkl\"\n",
    "    pickle.dump(parsed_data, open(fname, \"wb\"), protocol=pickle.HIGHEST_PROTOCOL)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dd022511-e165-4354-9c73-fe540332205e",
   "metadata": {},
   "source": [
    "Reload and retest. Ensure there are no zwitterions..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4e81ea9f-aa0f-43b3-8ce2-4c0261ad1116",
   "metadata": {},
   "outputs": [],
   "source": [
    "for file in Path(\"data/qm9/ml_ready/random_splits/\").iterdir():\n",
    "    loaded = pickle.load(open(file, \"rb\"))\n",
    "    for split in [\"test\", \"train\", \"val\"]:\n",
    "        assert np.sum([\"+\" in smi for smi in loaded[split][\"origin_smiles\"]]) == 0\n",
    "        assert np.sum([\"-\" in smi for smi in loaded[split][\"origin_smiles\"]]) == 0"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3d88b23c-b90d-4b61-ac9b-bddbd870ffff",
   "metadata": {},
   "source": [
    "## Train on molecules with at most `N` atoms"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf783b12-6536-455b-b7ea-02294ecc4b24",
   "metadata": {},
   "source": [
    "Get a production-quality plot for demonstrating how the database scales with atom number/molecule."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "bc457737-ffb0-424d-97b7-a0f18dbf94ae",
   "metadata": {},
   "outputs": [],
   "source": [
    "atom_counts_sorted = dict()\n",
    "\n",
    "int_grid = np.arange(1, 10)\n",
    "\n",
    "ATOM = \"C\"\n",
    "tmp_data = data[ATOM]\n",
    "atom_counts = list(map(atom_count, tmp_data[\"origin_smiles\"]))\n",
    "atom_counts = np.array([sum(xx.values()) for xx in atom_counts])\n",
    "atom_counts_counter = Counter(atom_counts)\n",
    "atom_counts_sorted[\"C\"] = [atom_counts_counter[ii] for ii in int_grid]\n",
    "\n",
    "ATOM = \"N\"\n",
    "tmp_data = data[ATOM]\n",
    "atom_counts = list(map(atom_count, tmp_data[\"origin_smiles\"]))\n",
    "atom_counts = np.array([sum(xx.values()) for xx in atom_counts])\n",
    "atom_counts_counter = Counter(atom_counts)\n",
    "atom_counts_sorted[\"N\"] = [atom_counts_counter[ii] for ii in int_grid]\n",
    "\n",
    "ATOM = \"O\"\n",
    "tmp_data = data[ATOM]\n",
    "atom_counts = list(map(atom_count, tmp_data[\"origin_smiles\"]))\n",
    "atom_counts = np.array([sum(xx.values()) for xx in atom_counts])\n",
    "atom_counts_counter = Counter(atom_counts)\n",
    "atom_counts_sorted[\"O\"] = [atom_counts_counter[ii] for ii in int_grid]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "b20d7c23-93e1-4931-8656-98eeb53a474f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from matplotlib.ticker import AutoMinorLocator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "250beda0-83ef-458d-b360-343787d383c6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 750x375 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create a well-balanced testing set\n",
    "fig, ax = plt.subplots(1, 1, figsize=(3, 1.5))\n",
    "\n",
    "\n",
    "ax.plot(int_grid, np.log10(atom_counts_sorted[\"C\"]), 'ko-', label=r\"C\")\n",
    "ax.plot(int_grid, np.log10(atom_counts_sorted[\"N\"]), 'bo-', label=r\"N\")\n",
    "ax.plot(int_grid, np.log10(atom_counts_sorted[\"O\"]), 'ro-', label=r\"O\")\n",
    "\n",
    "\n",
    "adj.set_grids(ax, grid=False)\n",
    "ax.set_xticks(int_grid)\n",
    "ax.tick_params(axis=\"x\", which=\"minor\", bottom=False, top=False)\n",
    "ax.tick_params(axis=\"y\", which=\"minor\", left=True, right=True)\n",
    "\n",
    "ax.set_ylabel(r\"$\\log_{10} \\mathrm{Counts}$\")\n",
    "ax.set_xlabel(r\"Heavy atoms/molecule\")\n",
    "\n",
    "ax.legend(frameon=False)\n",
    "\n",
    "ax.set_yticks([0, 2, 4, 6])\n",
    "\n",
    "minor_locator = AutoMinorLocator(2)\n",
    "ax.yaxis.set_minor_locator(minor_locator)\n",
    "\n",
    "adj.set_ylim(ax, 0, 6)\n",
    "\n",
    "\n",
    "plt.savefig(\"Figures/qm9_N-2.svg\", bbox_inches=\"tight\", dpi=300)\n",
    "# plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dd8c2431-7670-4e61-9c44-5668d21a3bde",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "for max_training_atoms_per_molecule in range(5, 9):\n",
    "    for CENTRAL_ATOM in [\"C\", \"N\", \"O\"]:\n",
    "        print(CENTRAL_ATOM)\n",
    "        parsed_data = split_qm9_data_by_number_of_total_atoms(\n",
    "            data[CENTRAL_ATOM],\n",
    "            max_training_atoms_per_molecule=max_training_atoms_per_molecule,\n",
    "            test_atoms_per_molecule=9,\n",
    "            prop_val=0.1,\n",
    "            seed=123\n",
    "        )\n",
    "        now = datetime.now().strftime(\"%y%m%d\")\n",
    "        fname = f\"data/qm9/ml_ready/XANES-{now}-ACSF-{CENTRAL_ATOM}-TRAIN-ATMOST-{max_training_atoms_per_molecule}-TOTAL-ATOMS.pkl\"\n",
    "        print(fname)\n",
    "        pickle.dump(parsed_data, open(fname, \"wb\"), protocol=pickle.HIGHEST_PROTOCOL)\n",
    "\n",
    "        parsed_data, max_column = pca_decomposition_inputs(parsed_data)\n",
    "        fname = f\"data/qm9/ml_ready/XANES-{now}-ACSF-{CENTRAL_ATOM}-TRAIN-ATMOST-{max_training_atoms_per_molecule}-TOTAL-ATOMS-PCA-decomp-maxcol-{max_column}.pkl\"\n",
    "        print(fname)\n",
    "        pickle.dump(parsed_data, open(fname, \"wb\"), protocol=pickle.HIGHEST_PROTOCOL)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a60d9c1a-9d5b-4b66-9a9d-e1b043043bbf",
   "metadata": {},
   "source": [
    "Reload and retest. Ensure there are no zwitterions..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4d746e1c-e117-4989-b6fe-6ae7a3f7d420",
   "metadata": {},
   "outputs": [],
   "source": [
    "for file in Path(\"data/qm9/ml_ready/by_total_atoms/\").iterdir():\n",
    "    loaded = pickle.load(open(file, \"rb\"))\n",
    "    for split in [\"test\", \"train\", \"val\"]:\n",
    "        assert np.sum([\"+\" in smi for smi in loaded[split][\"origin_smiles\"]]) == 0\n",
    "        assert np.sum([\"-\" in smi for smi in loaded[split][\"origin_smiles\"]]) == 0"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.12"
  },
  "toc-autonumbering": true,
  "toc-showmarkdowntxt": false
 },
 "nbformat": 4,
 "nbformat_minor": 5
}