Skip to content

Instantly share code, notes, and snippets.

@artyomLisovskij

artyomLisovskij/ai.ipynb

Created July 12, 2018 07:29
Show Gist options
  • Save artyomLisovskij/ced27c34b514b1017f360f760d3782c8 to your computer and use it in GitHub Desktop.
Save artyomLisovskij/ced27c34b514b1017f360f760d3782c8 to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
ai.ipynb
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Задача:\n",
"1. матрица (a1 b1 c1, a2 b2, c2, a3 b3 c3), найти детерминант матрицы для входящих параметров a1 b1 c1 a2 b2 c2 a3 b3 c3"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Для начала создадим обучающую выбурку и сохраним в файл формата csv."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Процедура для работы с файлом будет принимать множество данных, названия полей и имя файла."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import csv\n",
" \n",
"def writeCSV(fileName, data, fieldNames):\n",
" with open(fileName, 'w', newline='') as file:\n",
" writer = csv.DictWriter(file, fieldnames=fieldNames)\n",
" \n",
" writer.writeheader()\n",
" for i in data:\n",
" writer.writerow(i)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Функция для чтения данных."
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"def readCSV(fileName):\n",
" with open(fileName) as file:\n",
" reader = csv.DictReader(file)\n",
" for row in reader:\n",
" print(row)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Необходимо создать функцию которая будет генерировать матрицу и ответ к ней. \n",
"Матрица будет задаваться массивом 3x3."
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"import random\n",
"import numpy as np\n",
"from numpy.linalg import det as det\n",
"from numpy import array\n",
"\n",
"random.seed()\n",
"\n",
"def genMatrixWithDet():\n",
" minel = random.randint(-1000, 1000)\n",
" maxel = random.randint(-1000, 1000)\n",
" if minel > maxel:\n",
" minel, maxel = maxel, minel\n",
" matrix = array([[random.randint(minel, maxel) for v in range(3)] for val in range(3)])\n",
" ans = int(det(matrix))\n",
" \n",
" return matrix, ans"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(array([[-123, -59, -20],\n",
" [ 12, -125, -15],\n",
" [ -66, -116, -41]]), -310952)"
]
},
"execution_count": 4,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"genMatrixWithDet()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Заполним файл данными"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"titles = ['a1', 'b1', 'c1', 'a2', 'b2', 'c2', 'a3', 'b3','c3', 'det']\n",
"data = []\n",
"\n",
"for i in range(10000):\n",
" m, ans = genMatrixWithDet();\n",
" data.append({x : y for x, y in zip(titles, np.append(m.flatten(), ans))})\n",
" \n",
"writeCSV('train.csv', data, titles)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Загрузим данные"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>a1</th>\n",
" <th>b1</th>\n",
" <th>c1</th>\n",
" <th>a2</th>\n",
" <th>b2</th>\n",
" <th>c2</th>\n",
" <th>a3</th>\n",
" <th>b3</th>\n",
" <th>c3</th>\n",
" <th>det</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>-622</td>\n",
" <td>-665</td>\n",
" <td>-281</td>\n",
" <td>-151</td>\n",
" <td>-666</td>\n",
" <td>-239</td>\n",
" <td>-943</td>\n",
" <td>-205</td>\n",
" <td>-308</td>\n",
" <td>-48282287</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>496</td>\n",
" <td>202</td>\n",
" <td>404</td>\n",
" <td>161</td>\n",
" <td>426</td>\n",
" <td>384</td>\n",
" <td>676</td>\n",
" <td>492</td>\n",
" <td>598</td>\n",
" <td>-18706123</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>473</td>\n",
" <td>562</td>\n",
" <td>541</td>\n",
" <td>453</td>\n",
" <td>483</td>\n",
" <td>597</td>\n",
" <td>397</td>\n",
" <td>306</td>\n",
" <td>604</td>\n",
" <td>2264810</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>-769</td>\n",
" <td>-831</td>\n",
" <td>-798</td>\n",
" <td>-802</td>\n",
" <td>-741</td>\n",
" <td>-771</td>\n",
" <td>-818</td>\n",
" <td>-739</td>\n",
" <td>-700</td>\n",
" <td>-7556876</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>318</td>\n",
" <td>411</td>\n",
" <td>45</td>\n",
" <td>-37</td>\n",
" <td>428</td>\n",
" <td>14</td>\n",
" <td>-760</td>\n",
" <td>-491</td>\n",
" <td>-782</td>\n",
" <td>-105057195</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" a1 b1 c1 a2 b2 c2 a3 b3 c3 det\n",
"0 -622 -665 -281 -151 -666 -239 -943 -205 -308 -48282287\n",
"1 496 202 404 161 426 384 676 492 598 -18706123\n",
"2 473 562 541 453 483 597 397 306 604 2264810\n",
"3 -769 -831 -798 -802 -741 -771 -818 -739 -700 -7556876\n",
"4 318 411 45 -37 428 14 -760 -491 -782 -105057195"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from pandas import read_csv\n",
"\n",
"dataset = read_csv('train.csv', ',')\n",
"dataset.head()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Далее следует нормировать данные"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>a1</th>\n",
" <th>b1</th>\n",
" <th>c1</th>\n",
" <th>a2</th>\n",
" <th>b2</th>\n",
" <th>c2</th>\n",
" <th>a3</th>\n",
" <th>b3</th>\n",
" <th>c3</th>\n",
" <th>det</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>-0.624937</td>\n",
" <td>-0.668514</td>\n",
" <td>-0.281485</td>\n",
" <td>-0.146747</td>\n",
" <td>-0.669014</td>\n",
" <td>-0.242348</td>\n",
" <td>-0.949546</td>\n",
" <td>-0.205258</td>\n",
" <td>-0.312563</td>\n",
" <td>-0.222728</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>0.499246</td>\n",
" <td>0.205038</td>\n",
" <td>0.405921</td>\n",
" <td>0.167927</td>\n",
" <td>0.429577</td>\n",
" <td>0.382840</td>\n",
" <td>0.684157</td>\n",
" <td>0.499494</td>\n",
" <td>0.597990</td>\n",
" <td>-0.199345</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>0.476119</td>\n",
" <td>0.567758</td>\n",
" <td>0.543402</td>\n",
" <td>0.462431</td>\n",
" <td>0.486922</td>\n",
" <td>0.596588</td>\n",
" <td>0.402624</td>\n",
" <td>0.311426</td>\n",
" <td>0.604020</td>\n",
" <td>-0.182765</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>-0.772750</td>\n",
" <td>-0.835768</td>\n",
" <td>-0.800301</td>\n",
" <td>-0.803328</td>\n",
" <td>-0.744467</td>\n",
" <td>-0.776217</td>\n",
" <td>-0.823411</td>\n",
" <td>-0.745197</td>\n",
" <td>-0.706533</td>\n",
" <td>-0.190530</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>0.320261</td>\n",
" <td>0.415617</td>\n",
" <td>0.045660</td>\n",
" <td>-0.031770</td>\n",
" <td>0.431590</td>\n",
" <td>0.011540</td>\n",
" <td>-0.764884</td>\n",
" <td>-0.494439</td>\n",
" <td>-0.788945</td>\n",
" <td>-0.267615</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" a1 b1 c1 a2 b2 c2 a3 \\\n",
"0 -0.624937 -0.668514 -0.281485 -0.146747 -0.669014 -0.242348 -0.949546 \n",
"1 0.499246 0.205038 0.405921 0.167927 0.429577 0.382840 0.684157 \n",
"2 0.476119 0.567758 0.543402 0.462431 0.486922 0.596588 0.402624 \n",
"3 -0.772750 -0.835768 -0.800301 -0.803328 -0.744467 -0.776217 -0.823411 \n",
"4 0.320261 0.415617 0.045660 -0.031770 0.431590 0.011540 -0.764884 \n",
"\n",
" b3 c3 det \n",
"0 -0.205258 -0.312563 -0.222728 \n",
"1 0.499494 0.597990 -0.199345 \n",
"2 0.311426 0.604020 -0.182765 \n",
"3 -0.745197 -0.706533 -0.190530 \n",
"4 -0.494439 -0.788945 -0.267615 "
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from sklearn import preprocessing\n",
"from pandas import DataFrame\n",
"\n",
"x = dataset.values.astype(float)\n",
"min_max_scaler = preprocessing.MinMaxScaler(feature_range = (-1, 1))\n",
"x_scaled = min_max_scaler.fit_transform(x)\n",
"dataset = DataFrame(x_scaled, columns=titles)\n",
"dataset.head()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Составил тренировочный набор и тестовый"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"from sklearn.model_selection import train_test_split\n",
"\n",
"trg = dataset['det']\n",
"trn = dataset.drop('det', axis=1)\n",
"\n",
"Xtrn, Xtest, Ytrn, Ytest = train_test_split(trn, trg, test_size=0.3)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Будем использовать нейронную сеть так как зависимость нелинейная"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {
"scrolled": true
},
"outputs": [
{
"data": {
"text/plain": [
"MLPRegressor(activation='tanh', alpha=0.001, batch_size='auto', beta_1=0.9,\n",
" beta_2=0.999, early_stopping=False, epsilon=1e-08,\n",
" hidden_layer_sizes=(50, 9, 9), learning_rate='constant',\n",
" learning_rate_init=0.001, max_iter=10000, momentum=0.9,\n",
" nesterovs_momentum=True, power_t=0.5, random_state=None,\n",
" shuffle=True, solver='lbfgs', tol=0.0001, validation_fraction=0.1,\n",
" verbose=False, warm_start=False)"
]
},
"execution_count": 9,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from sklearn.neural_network import MLPRegressor\n",
"\n",
"model = MLPRegressor(hidden_layer_sizes=(50, 9, 9), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.001)\n",
"model.fit(Xtrn, Ytrn)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Провем точность на тренировочных данных"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"0.966348868203\n",
"0.00819768036411\n"
]
}
],
"source": [
"from sklearn.metrics import r2_score\n",
"from sklearn.metrics import mean_absolute_error\n",
"\n",
"print(r2_score(Ytest, model.predict(Xtest)))\n",
"print(mean_absolute_error(Ytest, model.predict(Xtest)))"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [],
"source": [
"# hidden_layer_sizes=(100, 100, 100), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.001 R2=0.966 trn=7000 test=3000\n",
"#hidden_layer_sizes=(100, 100, 100), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.971 trn=7000 test=3000\n",
"#hidden_layer_sizes=(9), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.628 trn=7000 test=3000\n",
"#hidden_layer_sizes=(81), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.903 trn=7000 test=3000\n",
"#hidden_layer_sizes=(100), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.909 trn=7000 test=3000\n",
"#hidden_layer_sizes=(9, 9), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.80 trn=7000 test=3000\n",
"#hidden_layer_sizes=(9, 9, 9), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.634 trn=7000 test=3000\n",
"#hidden_layer_sizes=(50, 9), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.939 trn=7000 test=3000\n",
"#hidden_layer_sizes=(50, 40), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.947 trn=7000 test=3000\n",
"#hidden_layer_sizes=(81, 9), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.944 trn=7000 test=3000\n",
"#hidden_layer_sizes=(81, 30), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.948 trn=7000 test=3000\n",
"#hidden_layer_sizes=(81, 81), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.947 trn=7000 test=3000\n",
"#hidden_layer_sizes=(50, 9, 9), max_iter=10000, activation='tanh', solver='lbfgs', alpha=0.01 R2=0.968 trn=7000 test=3000"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Проверим результат на своем примере"
]
},
{
"cell_type": "code",
"execution_count": 31,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[[695 751 309]\n",
" [598 902 865]\n",
" [685 893 585]]\n",
"-13766184\n",
"\n",
"expected = -0.195439473202\n",
"ans = [-0.19358827] \n",
"\n",
"\n",
"after inverse_transform\n",
"\n",
"expected = -13766184\n",
"ans = -11424692\n"
]
}
],
"source": [
"m, a = genMatrixWithDet()\n",
"print(m)\n",
"print(a)\n",
"tmp = np.append(m.flatten(), a)\n",
"\n",
"x_y = min_max_scaler.transform(tmp.flatten().reshape(1, -1).astype(float))\n",
"x = np.delete(x_y, 9).reshape(1, -1)\n",
"\n",
"print('\\nexpected = ', x_y[0][9])\n",
"print('ans = ', model.predict(x), '\\n')\n",
"\n",
"print('\\nafter inverse_transform\\n')\n",
"\n",
"print('expected = ', a)\n",
"print('ans = ', int(min_max_scaler.inverse_transform(np.append(x, model.predict(x)).reshape(1, -1))[0][9]))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"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": "3.6.6"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment