How to Develop a 1D Generative Adversarial Network From Scratch in PyTorch (Part 1)

h1ros

Jul 2, 2019, 6:27:49 AM

Goal¶

This post is inspired by the blog "Machine Learning Mastery - How to Develop a 1D Generative Adversarial Network From Scratch in Keras" written by Jason Brownlee, PhD. But to learn step-by-step, I will describe the same concept with PyTorch.

This post will cover the followings:

Part 1:

Select a One-Dimensional Function
Define a Discriminator Model

Reference

Libraries¶

In [33]:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

# PyTorch
import torch
from torch import nn
from torch import optim
from torchviz import make_dot

Create a target 1-D function¶

In [3]:

def f(x):
    return x **2

In [17]:

n = 100
sigma = 10
x = sigma * (np.random.random(size=n) - 0.5)
plt.plot(x, f(x), '.');
plt.title('Target function $f(x)$');
plt.xlabel('randomly sampled x');
plt.ylabel('$f(x)$');

Define a Discriminator Model¶

The definition of a discriminator model is that it will classify the input data into real or fake

In [35]:

# Build a feed-forward network
model = nn.Sequential(nn.Linear(2, 25),
                      nn.ReLU(),
                      nn.Sigmoid()
                     )

# Loss
criterion = nn.CrossEntropyLoss()

# Optimizer
optimizer = optim.Adam(model.parameters(), lr=0.001)

In [36]:

# Visualize this neural network
x = torch.zeros(0, 2, dtype=torch.float, requires_grad=False)
out = model(x)
make_dot(out)

Out[36]:

Create real and fake samples¶

In [79]:

def generate_samples(size=100, label='real'):
    """Generate samples with real or fake label
    """
    x = np.random.randn(size, 1)
    x2 = f(x)
    
    y = np.ones((size, 1)) * (label == 'real')
    return np.hstack([x, x2]), y

In [80]:

X, y = generate_samples()

In [82]:

x[:5]

Out[82]:

array([[ 1.2483621 ,  1.55840793],
       [ 0.57980381,  0.33617245],
       [-0.06718955,  0.00451444],
       [-1.95352245,  3.81624995],
       [-1.14922801,  1.32072501]])

In [83]:

y[:5]

Out[83]:

array([[1.],
       [1.],
       [1.],
       [1.],
       [1.]])

Goal¶

Libraries¶

Create a target 1-D function¶

Define a Discriminator Model¶

Create real and fake samples¶

Comments