Predictors/train_predictor.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "import re\n",
    "import csv\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from keras.models import Sequential,load_model\n",
    "from keras import optimizers, callbacks\n",
    "from keras.layers import Dense, Dropout\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "element_list = ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr','Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu']\n",
    "df_train = pd.read_csv('../Data/oqmd_train.csv')\n",
    "df_test = pd.read_csv('../Data/oqmd_test.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convert string representation of molecular formula to one-hot encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def counts2vector(pairs):\n",
    "    vec = [0]*len(element_list)\n",
    "    for pair in pairs:\n",
    "        vec[element_list.index(pair[0])]+=int(pair[1])\n",
    "    vec = np.array(vec)\n",
    "    return vec\n",
    "\n",
    "def onehot(enc):\n",
    "    rep = np.zeros((len(element_list),11))\n",
    "    rep[:,0] = 1\n",
    "    for i,val in enumerate(enc):\n",
    "        if val!=0:\n",
    "            rep[i][0]=0\n",
    "            rep[i][val]=1\n",
    "    return rep.flatten()\n",
    "\n",
    "def encode(df):\n",
    "    formulae = df[['comp']]\n",
    "    counts = pd.Series(formulae.values.flatten()).str.findall(r\"([a-z]+)([0-9]+)\", re.I)\n",
    "    df['encoding'] = counts.apply(counts2vector)\n",
    "    df['onehot'] = df['encoding'].apply(onehot)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "encode(df_train)\n",
    "encode(df_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Extract the encoding and target values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = df_train.onehot.values\n",
    "X_train = np.stack(X_train,axis=0)\n",
    "X_test = df_test.onehot.values\n",
    "X_test = np.stack(X_test,axis=0)\n",
    "y_train = df_train[['delta_e','volume_pa','energy_pa']].values\n",
    "y_test = df_test[['delta_e','volume_pa','energy_pa']].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create the layers of the neural network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def make_model(dropout=True):\n",
    "    \n",
    "    model = Sequential()\n",
    "    model.add(Dense(1024,input_dim=979,activation='relu'))\n",
    "    model.add(Dense(1024,activation='relu'))\n",
    "    \n",
    "    if dropout:\n",
    "        model.add(Dropout(0.7))\n",
    "    model.add(Dense(1024,activation='relu'))\n",
    "    if dropout:\n",
    "        model.add(Dropout(0.7))\n",
    "    model.add(Dense(512,activation='relu'))\n",
    "    if dropout:\n",
    "        model.add(Dropout(0.5))\n",
    "    \n",
    "    model.add(Dense(128,activation='relu'))\n",
    "    model.add(Dense(64,activation='relu'))\n",
    "    model.add(Dense(32,activation='relu'))\n",
    "    model.add(Dense(1,activation='linear'))\n",
    "\n",
    "    model.compile(loss='mean_absolute_error',optimizer=optimizers.adam())\n",
    "    model.summary()\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Formation Enthalpy Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train with Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = make_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reduce_lr = callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1,\n",
    "                              patience=5, min_lr=0.00001,verbose=1)\n",
    "checkpoint = callbacks.ModelCheckpoint(filepath='delta_e_intermediate.h5', monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "history = model.fit(X_train, y_train[:,0], epochs=120, validation_split=0.2,batch_size=256,callbacks=[reduce_lr,checkpoint])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Retrain without dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = make_model(dropout=False)\n",
    "model.load_weights('delta_e_intermediate.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reduce_lr = callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1,\n",
    "                              patience=5, min_lr=0.00001,verbose=1)\n",
    "checkpoint = callbacks.ModelCheckpoint(filepath='delta_e_best_model.h5', monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "history = model.fit(X_train, y_train[:,0], epochs=120, validation_split=0.2,batch_size=256,callbacks=[reduce_lr,checkpoint])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.evaluate(X_test,y_test[:,0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Volume Per Atom Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train with Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = make_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reduce_lr = callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1,\n",
    "                              patience=5, min_lr=0.00001,verbose=1)\n",
    "checkpoint = callbacks.ModelCheckpoint(filepath='volume_pa_intermediate.h5', monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "history = model.fit(X_train, y_train[:,1], epochs=120, validation_split=0.2,batch_size=256,callbacks=[reduce_lr,checkpoint])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Retrain without dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = make_model(dropout=False)\n",
    "model.load_weights('volume_pa_intermediate.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reduce_lr = callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1,\n",
    "                              patience=5, min_lr=0.00001,verbose=1)\n",
    "checkpoint = callbacks.ModelCheckpoint(filepath='volume_pa_best_model.h5', monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "history = model.fit(X_train, y_train[:,1], epochs=120, validation_split=0.2,batch_size=256,callbacks=[reduce_lr,checkpoint])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.evaluate(X_test,y_test[:,1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Energy Per Atom Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train with Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = make_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reduce_lr = callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1,\n",
    "                              patience=5, min_lr=0.00001,verbose=1)\n",
    "checkpoint = callbacks.ModelCheckpoint(filepath='energy_pa_intermediate.h5', monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "history = model.fit(X_train, y_train[:,2], epochs=120, validation_split=0.2,batch_size=256,callbacks=[reduce_lr,checkpoint])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Retrain without dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = make_model(dropout=False)\n",
    "model.load_weights('energy_pa_intermediate.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "reduce_lr = callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1,\n",
    "                              patience=5, min_lr=0.00001,verbose=1)\n",
    "checkpoint = callbacks.ModelCheckpoint(filepath='energy_pa_best_model.h5', monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "history = model.fit(X_train, y_train[:,2], epochs=120, validation_split=0.2,batch_size=256,callbacks=[reduce_lr,checkpoint])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.evaluate(X_test,y_test[:,2])"
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}