Skip to content

Instantly share code, notes, and snippets.

@NTT123

NTT123/mixture-density-network.ipynb

Created October 11, 2018 04:57
Show Gist options
  • Save NTT123/618d3ae7ea15506d1396691ee10dcdf5 to your computer and use it in GitHub Desktop.
Save NTT123/618d3ae7ea15506d1396691ee10dcdf5 to your computer and use it in GitHub Desktop.
Download ZIP
Mixture Density Network.ipynb
Display the source blob
Display the rendered blob
Raw
mixture-density-network.ipynb
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"name": "Mixture Density Network.ipynb",
"version": "0.3.2",
"provenance": [],
"include_colab_link": true
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
}
},
"cells": [
{
"cell_type": "markdown",
"metadata": {
"id": "view-in-github",
"colab_type": "text"
},
"source": [
"[View in Colaboratory](https://colab.research.google.com/gist/NTT123/618d3ae7ea15506d1396691ee10dcdf5/mixture-density-network.ipynb)"
]
},
{
"metadata": {
"id": "XrsUUHizruIm",
"colab_type": "text"
},
"cell_type": "markdown",
"source": [
"# 2D Mixture Density Network!"
]
},
{
"metadata": {
"id": "dS8brYDAro5z",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"# http://pytorch.org/\n",
"from os import path\n",
"from wheel.pep425tags import get_abbr_impl, get_impl_ver, get_abi_tag\n",
"platform = '{}{}-{}'.format(get_abbr_impl(), get_impl_ver(), get_abi_tag())\n",
"\n",
"accelerator = 'cu80' if path.exists('/opt/bin/nvidia-smi') else 'cpu'\n",
"\n",
"!pip install -q http://download.pytorch.org/whl/{accelerator}/torch-0.4.1-{platform}-linux_x86_64.whl torchvision\n",
"import torch\n",
"import torch.nn as nn\n",
"device = torch.device(\"cpu\")\n",
"import math\n",
"import numpy as np\n",
"\n",
"import matplotlib.pyplot as plt\n",
"import matplotlib.colors as mcolors\n",
"from pylab import rcParams\n",
"\n",
"plt.style.use('ggplot')\n",
"rcParams['figure.figsize'] = 3,3"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "Iho8bIrtr8DN",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"# 10 Gaussian mixtures\n",
"M = 10\n",
"weights = (0.2*torch.rand(M)).softmax(dim=0)\n",
"means = 20* (torch.rand(M, 2) - 0.5)\n",
"covs = [ torch.randn(2,2) for _ in range(M) ]\n",
"covs = [ torch.matmul(A, A.t()) + 0.1*torch.eye(2) for A in covs ]\n",
"Zs = [ torch.distributions.MultivariateNormal(loc=means[c], \n",
" covariance_matrix=covs[c]) for c in range(M) ]\n"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "1iiMtqvCA03D",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"# Generate data\n",
"N = 100000\n",
"choices = np.random.choice(range(M), size=N, p=weights.numpy())\n",
"D = torch.empty( (N, 2) )\n",
"\n",
"for id, c in enumerate(choices): \n",
" D[id,: ] = Zs[c].sample()"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "E-2dQjFMHfTX",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"# zero mean, 1 std\n",
"D = (D - D.mean(dim=0)) / D.std(dim=0)"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "zrs8N6UrITsa",
"colab_type": "code",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
},
"outputId": "743de01a-6f63-4c8a-984f-a62a1dc29bc2"
},
"cell_type": "code",
"source": [
"D.std(dim=0)"
],
"execution_count": 5,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"tensor([1., 1.])"
]
},
"metadata": {
"tags": []
},
"execution_count": 5
}
]
},
{
"metadata": {
"id": "mC5AoMj7vDDM",
"colab_type": "code",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 215
},
"outputId": "6c72bcd7-3f0a-4416-c55f-a6d9abf46ef4"
},
"cell_type": "code",
"source": [
"\n",
"\n",
"\n",
"_ = plt.hist2d(D[:, 0].numpy(), D[:, 1].numpy(), bins=100, range=[[-3, 3],[-3,3]])"
],
"execution_count": 6,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAMgAAADGCAYAAACXUs/uAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
"text/plain": [
"<matplotlib.figure.Figure at 0x7f007f2517b8>"
]
},
"metadata": {
"tags": []
}
}
]
},
{
"metadata": {
"id": "sDs1xFODJO1u",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"class MDN(nn.Module):\n",
" def __init__(self, m):\n",
" super().__init__()\n",
" \n",
" self.fc1 = nn.Linear(2, 100)\n",
" self.selu1 = nn.Tanh()\n",
" self.m = m\n",
" \n",
" s = 6 * m # weight, meanx, meany, varx, vary rho\n",
" self.fc2 = nn.Linear(100, s)\n",
" \n",
" def forward(self, x):\n",
" x = self.selu1(self.fc1(x))\n",
" out = self.fc2(x)\n",
" \n",
" return out\n"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "3HciA4LwKvxk",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"m = 40\n",
"model = MDN(m).to(device) # 20 mixtures"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "BP7U8pi9K0Hg",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"py"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "w1u0wdOPK-Ol",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"inp = torch.randn(10, 2).to(device)\n",
"out = model(inp)"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "28xTSSPYOJjg",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"target = torch.randn(10,2).to(device)"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "RKl3hPGXLEtL",
"colab_type": "code",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
},
"outputId": "05b14ec8-9c9e-4816-ab68-44576b3c3e60"
},
"cell_type": "code",
"source": [
"myloss(out, target)"
],
"execution_count": 31,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"tensor(2.6001, grad_fn=<NegBackward>)"
]
},
"metadata": {
"tags": []
},
"execution_count": 31
}
]
},
{
"metadata": {
"id": "KbZ5klceQfqa",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
""
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "RoQBSJ_cLWix",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"from torch.utils.data import DataLoader\n",
"from torch.utils.data.dataset import Dataset\n",
"\n",
"class Mydataset(Dataset):\n",
"\n",
" def __init__(self, dt):\n",
" self.mydt = dt\n",
"\n",
" def __getitem__(self, index): \n",
" return self.mydt[index]\n",
"\n",
" def __len__(self):\n",
" return len(self.mydt)\n"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "8sALjPAoQzYu",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"optimizer = torch.optim.SGD(model.parameters(), lr=1e-1)\n",
"lr = torch.optim.lr_scheduler.ExponentialLR(optimizer, 0.8)"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "Ywvu2fJRQ6LU",
"colab_type": "code",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 2057
},
"outputId": "9e291393-2eb8-4a3d-d2a0-c68f878f3569"
},
"cell_type": "code",
"source": [
"model.train()\n",
"zer = torch.ones(32, 2).to(device).data\n",
"\n",
"for epoch in range(20):\n",
" train_loader = DataLoader( Mydataset(D), batch_size=32, shuffle=True)\n",
" lr.step()\n",
" me = 0.0\n",
" for batch_idx, data in enumerate(train_loader):\n",
" #print(data.size())\n",
" inp = data.to(device).data\n",
" output = model(zer)\n",
" loss = myloss(output, inp)\n",
" optimizer.zero_grad()\n",
" loss.backward()\n",
" optimizer.step()\n",
" \n",
" me = me + loss.item()\n",
" \n",
" if batch_idx % 500 == 499:\n",
" print(\"[%d : %d : %0.5f]: %0.5f\" % (epoch, batch_idx+1, lr.get_lr()[0], me / 500))\n",
" me = 0.0\n",
" \n",
" "
],
"execution_count": 42,
"outputs": [
{
"output_type": "stream",
"text": [
"[0 : 500 : 0.00115]: 1.63789\n",
"[0 : 1000 : 0.00115]: 1.64484\n",
"[0 : 1500 : 0.00115]: 1.62631\n",
"[0 : 2000 : 0.00115]: 1.63024\n",
"[0 : 2500 : 0.00115]: 1.63827\n",
"[0 : 3000 : 0.00115]: 1.64013\n",
"[1 : 500 : 0.00092]: 1.63638\n",
"[1 : 1000 : 0.00092]: 1.64615\n",
"[1 : 1500 : 0.00092]: 1.62593\n",
"[1 : 2000 : 0.00092]: 1.64217\n",
"[1 : 2500 : 0.00092]: 1.64123\n",
"[1 : 3000 : 0.00092]: 1.63096\n",
"[2 : 500 : 0.00074]: 1.63961\n",
"[2 : 1000 : 0.00074]: 1.62615\n",
"[2 : 1500 : 0.00074]: 1.64420\n",
"[2 : 2000 : 0.00074]: 1.63097\n",
"[2 : 2500 : 0.00074]: 1.63210\n",
"[2 : 3000 : 0.00074]: 1.64108\n",
"[3 : 500 : 0.00059]: 1.63530\n",
"[3 : 1000 : 0.00059]: 1.63679\n",
"[3 : 1500 : 0.00059]: 1.63642\n",
"[3 : 2000 : 0.00059]: 1.63991\n",
"[3 : 2500 : 0.00059]: 1.62682\n",
"[3 : 3000 : 0.00059]: 1.63724\n",
"[4 : 500 : 0.00047]: 1.63613\n",
"[4 : 1000 : 0.00047]: 1.62351\n",
"[4 : 1500 : 0.00047]: 1.63250\n",
"[4 : 2000 : 0.00047]: 1.65049\n",
"[4 : 2500 : 0.00047]: 1.63226\n",
"[4 : 3000 : 0.00047]: 1.63578\n",
"[5 : 500 : 0.00038]: 1.63465\n",
"[5 : 1000 : 0.00038]: 1.63242\n",
"[5 : 1500 : 0.00038]: 1.64190\n",
"[5 : 2000 : 0.00038]: 1.63781\n",
"[5 : 2500 : 0.00038]: 1.62324\n",
"[5 : 3000 : 0.00038]: 1.63544\n",
"[6 : 500 : 0.00030]: 1.62171\n",
"[6 : 1000 : 0.00030]: 1.63752\n",
"[6 : 1500 : 0.00030]: 1.63910\n",
"[6 : 2000 : 0.00030]: 1.63945\n",
"[6 : 2500 : 0.00030]: 1.62989\n",
"[6 : 3000 : 0.00030]: 1.64226\n",
"[7 : 500 : 0.00024]: 1.62444\n",
"[7 : 1000 : 0.00024]: 1.63705\n",
"[7 : 1500 : 0.00024]: 1.63977\n",
"[7 : 2000 : 0.00024]: 1.63628\n",
"[7 : 2500 : 0.00024]: 1.62043\n",
"[7 : 3000 : 0.00024]: 1.64010\n",
"[8 : 500 : 0.00019]: 1.62927\n",
"[8 : 1000 : 0.00019]: 1.63880\n",
"[8 : 1500 : 0.00019]: 1.63094\n",
"[8 : 2000 : 0.00019]: 1.63817\n",
"[8 : 2500 : 0.00019]: 1.63298\n",
"[8 : 3000 : 0.00019]: 1.63939\n",
"[9 : 500 : 0.00015]: 1.63136\n",
"[9 : 1000 : 0.00015]: 1.63009\n",
"[9 : 1500 : 0.00015]: 1.63269\n",
"[9 : 2000 : 0.00015]: 1.62601\n",
"[9 : 2500 : 0.00015]: 1.63500\n",
"[9 : 3000 : 0.00015]: 1.64604\n",
"[10 : 500 : 0.00012]: 1.62533\n",
"[10 : 1000 : 0.00012]: 1.63037\n",
"[10 : 1500 : 0.00012]: 1.64502\n",
"[10 : 2000 : 0.00012]: 1.62350\n",
"[10 : 2500 : 0.00012]: 1.63908\n",
"[10 : 3000 : 0.00012]: 1.64397\n",
"[11 : 500 : 0.00010]: 1.63478\n",
"[11 : 1000 : 0.00010]: 1.63547\n",
"[11 : 1500 : 0.00010]: 1.62571\n",
"[11 : 2000 : 0.00010]: 1.63491\n",
"[11 : 2500 : 0.00010]: 1.63768\n",
"[11 : 3000 : 0.00010]: 1.63386\n",
"[12 : 500 : 0.00008]: 1.64252\n",
"[12 : 1000 : 0.00008]: 1.62689\n",
"[12 : 1500 : 0.00008]: 1.64115\n",
"[12 : 2000 : 0.00008]: 1.63263\n",
"[12 : 2500 : 0.00008]: 1.62448\n",
"[12 : 3000 : 0.00008]: 1.63734\n",
"[13 : 500 : 0.00006]: 1.62581\n",
"[13 : 1000 : 0.00006]: 1.64151\n",
"[13 : 1500 : 0.00006]: 1.64488\n",
"[13 : 2000 : 0.00006]: 1.62632\n",
"[13 : 2500 : 0.00006]: 1.62985\n",
"[13 : 3000 : 0.00006]: 1.62826\n",
"[14 : 500 : 0.00005]: 1.64410\n",
"[14 : 1000 : 0.00005]: 1.62550\n",
"[14 : 1500 : 0.00005]: 1.62725\n",
"[14 : 2000 : 0.00005]: 1.64097\n",
"[14 : 2500 : 0.00005]: 1.63846\n",
"[14 : 3000 : 0.00005]: 1.62461\n",
"[15 : 500 : 0.00004]: 1.63615\n",
"[15 : 1000 : 0.00004]: 1.64376\n",
"[15 : 1500 : 0.00004]: 1.62264\n",
"[15 : 2000 : 0.00004]: 1.64407\n",
"[15 : 2500 : 0.00004]: 1.63888\n",
"[15 : 3000 : 0.00004]: 1.62125\n",
"[16 : 500 : 0.00003]: 1.63551\n",
"[16 : 1000 : 0.00003]: 1.61353\n",
"[16 : 1500 : 0.00003]: 1.64195\n",
"[16 : 2000 : 0.00003]: 1.65160\n",
"[16 : 2500 : 0.00003]: 1.63209\n",
"[16 : 3000 : 0.00003]: 1.62860\n",
"[17 : 500 : 0.00003]: 1.63436\n",
"[17 : 1000 : 0.00003]: 1.62711\n",
"[17 : 1500 : 0.00003]: 1.64353\n",
"[17 : 2000 : 0.00003]: 1.62368\n",
"[17 : 2500 : 0.00003]: 1.63820\n",
"[17 : 3000 : 0.00003]: 1.63200\n",
"[18 : 500 : 0.00002]: 1.63967\n",
"[18 : 1000 : 0.00002]: 1.63210\n",
"[18 : 1500 : 0.00002]: 1.62567\n",
"[18 : 2000 : 0.00002]: 1.63931\n",
"[18 : 2500 : 0.00002]: 1.63535\n",
"[18 : 3000 : 0.00002]: 1.63514\n",
"[19 : 500 : 0.00002]: 1.63749\n",
"[19 : 1000 : 0.00002]: 1.61178\n",
"[19 : 1500 : 0.00002]: 1.63418\n",
"[19 : 2000 : 0.00002]: 1.63456\n",
"[19 : 2500 : 0.00002]: 1.63777\n",
"[19 : 3000 : 0.00002]: 1.63995\n"
],
"name": "stdout"
}
]
},
{
"metadata": {
"id": "Gzy5cICTMa00",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"output = model( torch.ones(1, 2).to(device)).cpu()\n",
"w, mx, my, logvarx, logvary, rho = output.split(m, dim=1)\n",
"w = w.softmax(dim=1)\n",
"varx = logvarx.exp()\n",
"vary = logvary.exp()\n",
"rho = rho.tanh()\n"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "JFowKk3NVdWX",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"mx = mx.view(-1)\n",
"my = my.view(-1)\n",
"varx = varx.view(-1)\n",
"vary = vary.view(-1)\n",
"rho = rho.view(-1)\n",
"w = w.view(-1)"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "z8ztybkZV4Fm",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
"# Generate data\n",
"N = 100000\n",
"M=40\n",
"choices = np.random.choice(range(M), size=N, p=w.data.view(-1).numpy())\n",
"KK = torch.empty( (N, 2) )\n",
"\n",
"\n",
"means = torch.empty(M,2)\n",
"means[:,0] = mx\n",
"means[:,1] = my\n",
"\n",
"covs = [ torch.Tensor( [ [varx[i] , rho[i]*torch.sqrt(varx[i]*vary[i]) ], \n",
" [rho[i]*torch.sqrt(varx[i]*vary[i]), vary[i]] ] ) for i in range(M) ]\n",
"Zs = [ torch.distributions.MultivariateNormal(loc=means[c], \n",
" covariance_matrix=covs[c]) for c in range(M) ]\n",
"\n",
"for id, c in enumerate(choices): \n",
" KK[id,: ] = Zs[c].sample()"
],
"execution_count": 0,
"outputs": []
},
{
"metadata": {
"id": "m3KUwzTgXBfl",
"colab_type": "code",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 215
},
"outputId": "7a819661-6eed-4c5b-fd47-59ee229057c3"
},
"cell_type": "code",
"source": [
"plt.style.use('ggplot')\n",
"plt.minorticks_off()\n",
"rcParams['figure.figsize'] = 3,3\n",
"\n",
"\n",
"_ = plt.hist2d(KK[:, 0].numpy(), KK[:, 1].numpy(), bins=100 , range=[[-3, 3],[-3,3]])\n",
"rcParams['figure.figsize'] = 3,3\n"
],
"execution_count": 46,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAMgAAADGCAYAAACXUs/uAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFkRABMGCCIggWPg/z6DsqCIcwuQAAAAASUVORK5CYII=\n",
"text/plain": [
"<matplotlib.figure.Figure at 0x7f007c9087f0>"
]
},
"metadata": {
"tags": []
}
}
]
},
{
"metadata": {
"id": "NowyX10fvalC",
"colab_type": "code",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 215
},
"outputId": "82dfa6b4-e7df-44a9-e51c-4eacb7436e84"
},
"cell_type": "code",
"source": [
"plt.style.use('ggplot')\n",
"plt.minorticks_off()\n",
"rcParams['figure.figsize'] = 3,3\n",
"\n",
"\n",
"_ = plt.hist2d(D[:, 0].numpy(), D[:, 1].numpy(), bins=100, range=[[-3, 3],[-3,3]])"
],
"execution_count": 47,
"outputs": [
{
"output_type": "display_data",
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAMgAAADGCAYAAACXUs/uAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
"text/plain": [
"<matplotlib.figure.Figure at 0x7f007c8a2630>"
]
},
"metadata": {
"tags": []
}
}
]
},
{
"metadata": {
"id": "PVRLi_Tjg7IN",
"colab_type": "code",
"colab": {}
},
"cell_type": "code",
"source": [
""
],
"execution_count": 0,
"outputs": []
}
]
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment