{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Group-aware Sampling\n",
"In large dataset with groups, choice of the sampling method may depend on the number of groups and the group size. \n",
"\n",
"In this example we consider line plot where each line correspons to a group."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:26.908842Z",
"iopub.status.busy": "2024-04-17T07:30:26.908680Z",
"iopub.status.idle": "2024-04-17T07:30:27.222882Z",
"shell.execute_reply": "2024-04-17T07:30:27.222431Z"
}
},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" \n",
" "
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"import numpy as np\n",
"from lets_plot import *\n",
"\n",
"LetsPlot.setup_html()"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.236218Z",
"iopub.status.busy": "2024-04-17T07:30:27.236058Z",
"iopub.status.idle": "2024-04-17T07:30:27.238782Z",
"shell.execute_reply": "2024-04-17T07:30:27.238489Z"
}
},
"outputs": [],
"source": [
"def data(n_per_line, n_groups):\n",
" x_step = 2 * np.pi / (n_per_line - 1)\n",
" little_delta = x_step / 100\n",
" x_stops =np.arange(-np.pi, np.pi + little_delta, x_step)\n",
"\n",
" y_min, y_max = 1, 10\n",
" y_step = (y_max - y_min) / (n_groups - 1)\n",
" little_delta = y_step / 100\n",
" y_multiplier = np.arange(y_min, y_max + little_delta, y_step)\n",
"\n",
" x = []\n",
" y = []\n",
" c = []\n",
" for i in range(n_groups):\n",
" x.extend(x_stops)\n",
" y.extend([np.sin(x) * y_multiplier[i] for x in x_stops])\n",
" c.extend([str(i) for _ in x_stops])\n",
"\n",
" return dict(x = x, y = y, cond = c)"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.239851Z",
"iopub.status.busy": "2024-04-17T07:30:27.239715Z",
"iopub.status.idle": "2024-04-17T07:30:27.241398Z",
"shell.execute_reply": "2024-04-17T07:30:27.241055Z"
}
},
"outputs": [],
"source": [
"p = ggplot(mapping=aes('x','y',color='cond'))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 1. Data with large number of points per group but small number of groups."
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.242830Z",
"iopub.status.busy": "2024-04-17T07:30:27.242704Z",
"iopub.status.idle": "2024-04-17T07:30:27.248449Z",
"shell.execute_reply": "2024-04-17T07:30:27.248259Z"
}
},
"outputs": [],
"source": [
"n_per_line, n_groups = 1000, 10\n",
"dat = data(n_per_line, n_groups)"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.249755Z",
"iopub.status.busy": "2024-04-17T07:30:27.249617Z",
"iopub.status.idle": "2024-04-17T07:30:27.328808Z",
"shell.execute_reply": "2024-04-17T07:30:27.328600Z"
}
},
"outputs": [
{
"data": {
"text/html": [
" \n",
" "
],
"text/plain": [
""
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# The default 'systematic' sampling is fine in this case.\n",
"p + geom_line(data=dat)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 2. Data with small number of points per group but large number of groups."
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.332514Z",
"iopub.status.busy": "2024-04-17T07:30:27.332423Z",
"iopub.status.idle": "2024-04-17T07:30:27.348942Z",
"shell.execute_reply": "2024-04-17T07:30:27.348748Z"
}
},
"outputs": [],
"source": [
"n_per_line, n_groups = 30, 1000\n",
"dat = data(n_per_line, n_groups)"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.350478Z",
"iopub.status.busy": "2024-04-17T07:30:27.350395Z",
"iopub.status.idle": "2024-04-17T07:30:27.509492Z",
"shell.execute_reply": "2024-04-17T07:30:27.509264Z"
}
},
"outputs": [
{
"data": {
"text/html": [
" \n",
" "
],
"text/plain": [
""
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# The default systematic line sampling doesn't work very well in this case.\n",
"p + geom_line(data=dat)"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.517420Z",
"iopub.status.busy": "2024-04-17T07:30:27.517337Z",
"iopub.status.idle": "2024-04-17T07:30:27.586763Z",
"shell.execute_reply": "2024-04-17T07:30:27.586560Z"
}
},
"outputs": [
{
"data": {
"text/html": [
" \n",
" "
],
"text/plain": [
""
]
},
"execution_count": 8,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# Random group sampling works better.\n",
"p + geom_line(data=dat, sampling=sampling_group_random(10))"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.587914Z",
"iopub.status.busy": "2024-04-17T07:30:27.587828Z",
"iopub.status.idle": "2024-04-17T07:30:27.655528Z",
"shell.execute_reply": "2024-04-17T07:30:27.655330Z"
}
},
"outputs": [
{
"data": {
"text/html": [
" \n",
" "
],
"text/plain": [
""
]
},
"execution_count": 9,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# Systematic group sampling works even better.\n",
"p + geom_line(data=dat, sampling=sampling_group_systematic(10))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 3. Data with large number of points per group and large number of groups."
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.656704Z",
"iopub.status.busy": "2024-04-17T07:30:27.656588Z",
"iopub.status.idle": "2024-04-17T07:30:27.743363Z",
"shell.execute_reply": "2024-04-17T07:30:27.743053Z"
}
},
"outputs": [],
"source": [
"n_per_line, n_groups = 1000, 200\n",
"dat = data(n_per_line, n_groups)"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:27.744513Z",
"iopub.status.busy": "2024-04-17T07:30:27.744435Z",
"iopub.status.idle": "2024-04-17T07:30:28.290179Z",
"shell.execute_reply": "2024-04-17T07:30:28.289961Z"
}
},
"outputs": [
{
"data": {
"text/html": [
" \n",
" "
],
"text/plain": [
""
]
},
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"p + geom_line(data=dat)"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {
"execution": {
"iopub.execute_input": "2024-04-17T07:30:28.299529Z",
"iopub.status.busy": "2024-04-17T07:30:28.299441Z",
"iopub.status.idle": "2024-04-17T07:30:28.753506Z",
"shell.execute_reply": "2024-04-17T07:30:28.753303Z"
}
},
"outputs": [
{
"data": {
"text/html": [
" \n",
" "
],
"text/plain": [
""
]
},
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# A combination of points and group sampling works.\n",
"p + geom_line(data=dat, sampling=sampling_group_systematic(10)+sampling_systematic(200))"
]
}
],
"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": "3.10.13"
}
},
"nbformat": 4,
"nbformat_minor": 1
}