Generative Adversarial Networks: A Two-player game

Introduced in 2014 by Goodfellow. et al, Generative Adversarial Networks (GANs) revolutionized the field of Generative modeling. They proposed a new framework that generated very realistic synthetic data trained through a minimax two-player game.

With GANs, we don't explicitly learn the distribution of the data $p_{\text{data}}$, but we can still sample from it. Like VAEs, GANs also have two networks: a Generator and a Discriminator that are trained simultaneously.

A latent variable is sampled from a prior $\mathbf{z} \sim p(\mathbf{z})$ and passed through the Generator to obtain a fake sample $\mathbf{x} = G(\mathbf{z})$. Then the discriminator performs an image classification task that takes in all samples and classifies it as real ($1$) or fake ($0$). These are trained together such that while competing with each other, they make each other stronger at the same time.


To summarize,

  • A discriminator $D$ estimates the probability of a given sample coming from the real dataset. It works as a critic and is optimized to tell the fake samples from the real ones.

  • A generator $G$ outputs synthetic samples given a noise variable input $\mathbf{z}$. It is trained to capture the real data distribution (want $p_G \approx p_{\text{data}}$) so that its generative samples can be as real as possible, or in other words, can trick the discriminator to output higher probabilities.



Training Objective

We want to make sure that the Discriminator's decision over the real data is accurate, i.e. $D(\mathbf{x}) = 1$. This can be achieved by maximizing the log-likelihood over the data. \begin{equation*} \max_{\color{blue}{D}} \; \mathbb{E}_{\mathbf{x} \sim p_{\text{data}}} \left[ \log {\color{blue}{D}}(\mathbf{x}) \right] \tag{1} \end{equation*}

The Discriminator should also assign low probabilities to the fake samples generated by the Generator, i.e. $D(G(\mathbf{z})) = 0$. In other words, minimize the log-likelihood of the fake data. \begin{align*} \min_{\color{blue}{D}} \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} [\log({\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)] \tag{2.1} \end{align*} which can be re-written as, \begin{align*} & \max_{\color{blue}{D}} \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} [\log(1 -{\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)] \tag{2.2} \end{align*}


On the other hand, the Generator wants to fool the Discriminator such that it classifies fake samples as real, i.e. $D(G(\mathbf{z})) = 1$. This can be obtained by maximizing the log-likelihood over the fake data. \begin{align*} \max_{\color{orange}{G}} \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} [\log({\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)] \tag{3.1} \end{align*} which can be re-written as, \begin{align*} \min_{\color{orange}{G}} \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} [\log(1 -{\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)] \tag{3.2} \end{align*}


The full objective is given by,

\begin{align*} \large{\min_{\color{orange}{G}} \, \max_{\color{blue}{D}} \; \mathbb{E}_{\mathbf{x} \sim p_{\text{data}}} \left[ \log {\color{blue}{D}}(\mathbf{x}) \right] \; + \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} [\log(1 - {\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)]} \tag{4} \end{align*}

The objective of training GANs is therefore a minimax game: the generator $G$ is trying hard to trick the discriminator, while the critic model $D$ is trying hard not to be cheated. In this process, both networks force each other to improve their functionalities. And both are trained together with alternating gradient updates (gradient ascent on discriminator and gradient descent on generator).


However, at the start of training, the generator is really bad due to which $D(G(\mathbf{z}))$ is close to zero, creating a vanishing gradients problem for the Generator (as loss = 0). Therefore, instead of $\min_{\color{orange}{G}} \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} [\log(1 -{\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)]$, we use $\min_{\color{orange}{G}} \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} - [\log({\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)]$ or just $\max_{\color{orange}{G}} \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} [\log({\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)]$ (Equation 3.2) as an objective for the Generator.


Training Algorithm


Optimality

Let's write Equation 4 with a change of variables (ignoring the improvement of Generator for now), \begin{align*} & \min_{\color{orange}{G}} \, \max_{\color{blue}{D}} \; \mathbb{E}_{\mathbf{x} \sim p_{\text{data}}} \left[ \log {\color{blue}{D}}(\mathbf{x}) \right] \; + \; \mathbb{E}_{{\color{green}{\mathbf{z}}} \sim p({\color{green}{z}})} [\log(1 - {\color{blue}{D}} \left( {\color{orange}{G}}({\color{green}{z}})) \right)] \\ &= \min_{\color{orange}{G}} \, \max_{\color{blue}{D}} \; \mathbb{E}_{\mathbf{x} \sim p_{\text{data}}} \left[ \log {\color{blue}{D}}(\mathbf{x}) \right] \; + \; \mathbb{E}_{\mathbf{x} \sim p_{\color{orange}{G}} } [\log(1 - {\color{blue}{D}} (\mathbf{x}) )] \\ &= \min_{\color{orange}{G}} \, \int_{\mathbf{x}} \max_{\color{blue}{D}} \left( p_{\text{data}}(\mathbf{x}) \, \log {\color{blue}{D}}(\mathbf{x}) \; + \; p_{\color{orange}{G}}(\mathbf{x}) \, \log(1 - {\color{blue}{D}} (\mathbf{x}) ) \right)\\ \\ &\text{For the max function, we take the derivative wrt ${\color{blue}{D}}$ and assign it to zero.}\\ &\frac{p_{\text{data}}(\mathbf{x})}{{\color{blue}{D}}(\mathbf{x})} - \frac{p_{\color{orange}{G}}(\mathbf{x})}{(1 - {\color{blue}{D}} (\mathbf{x}))} = 0 \; \Rightarrow {{\color{blue}{D}}_{\max}(\mathbf{x})} = \frac{p_{\text{data}}(\mathbf{x})}{p_{\text{data}}(\mathbf{x}) + p_{\color{orange}{G}}(\mathbf{x})} \text{ [Optimal Discriminator]} \\ \\ &= \min_{\color{orange}{G}} \, \int_{\mathbf{x}} \left( p_{\text{data}}(\mathbf{x}) \, \log \frac{p_{\text{data}}(\mathbf{x})}{p_{\text{data}}(\mathbf{x}) + p_{\color{orange}{G}}(\mathbf{x})} \; + \; p_{\color{orange}{G}}(\mathbf{x}) \, \log \frac{p_{\color{orange}{G}}(\mathbf{x})}{p_{\text{data}}(\mathbf{x}) + p_{\color{orange}{G}}(\mathbf{x})} \right)\\ &= \min_{\color{orange}{G}} \, \left( \mathbf{E}_{\mathbf{x} \sim p_{\text{data}}} \, \left[ \log \frac{p_{\text{data}}(\mathbf{x})}{p_{\text{data}}(\mathbf{x}) + p_{\color{orange}{G}}(\mathbf{x})} \right] \; + \; \mathbf{E}_{\mathbf{x} \sim p_{\color{orange}{G}}} \, \left[ \log \frac{p_{\color{orange}{G}}(\mathbf{x})}{p_{\text{data}}(\mathbf{x}) + p_{\color{orange}{G}}(\mathbf{x})} \right] \right)\\ \\ &\text{Multiply and divide by 2} \\ &= \min_{\color{orange}{G}} \, \left( \mathbf{E}_{\mathbf{x} \sim p_{\text{data}}} \, \left[ \log \frac{2 \, p_{\text{data}}(\mathbf{x})}{p_{\text{data}}(\mathbf{x}) + p_{\color{orange}{G}}(\mathbf{x})} \right] \; + \; \mathbf{E}_{\mathbf{x} \sim p_{\color{orange}{G}}} \, \left[ \log \frac{2 \, p_{\color{orange}{G}}(\mathbf{x})}{p_{\text{data}}(\mathbf{x}) + p_{\color{orange}{G}}(\mathbf{x})} \right] - \log 4 \right) \\ \\ &\text{From the definition of KL-Divergence: } \mathbf{D}_{KL}(p || q) = \mathbf{E}_{x \sim p} \left[ \log \frac{p(x)}{q(x)} \right]\\ &= \min_{\color{orange}{G}} \, \left( \mathbf{D}_{KL} \left[ p_{\text{data}} \; || \; \frac{p_{\text{data}} + p_{\color{orange}{G}}}{2} \right] \, + \, \mathbf{D}_{KL} \left[ p_{\color{orange}{G}} \; || \; \frac{p_{\text{data}} + p_{\color{orange}{G}}}{2} \right] - \log 4 \right) \\ \\ &\text{From Jensen-Shannon divergence: } \mathbf{D}_{JS}(p || q) = \frac{1}{2} \mathbf{D}_{KL}(p || \frac{p + q}{2}) + \frac{1}{2} \mathbf{D}_{KL}(q || \frac{p + q}{2}) \\ &= \min_{\color{orange}{G}} \, \left( 2 \, \mathbf{D}_{JS} (p_{\text{data}} \, || \, p_{\color{orange}{G}}) - \log 4 \right)\\ \end{align*}

JS divergence is always non-negative and is zero only when the two distributions are equal. Hence, training a GAN with this objective, minimizes the JS divergence to zero, obtaining a global minimum when $p_{\text{data}} = p_{G}$.

The global minimum of the minimax game happens when, \begin{align*} \text{Optimal Generator: } & p_{\text{data}} = p_{G} \\ \text{Optimal Discriminator: } & D_{G}^*(\mathbf{x}) = \frac{p_{\text{data}}(\mathbf{x})}{p_{\text{data}}(\mathbf{x}) + p_G(\mathbf{x})} = \frac{1}{2} \end{align*}



Some prominent papers,

  1. Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks (DCGAN): Architectures of Generator and Discriminator created with CNNs.

  2. Wasserstein GAN (WGAN): Improvement over traditional GAN training.

  3. Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks (CycleGAN): For image-to-image translation

Comments

Post a Comment

Popular posts from this blog

The move_base ROS node

Image
move_base package provides the move_base node which is a major element of ROS Navigation stack. It moves the robot from its current position to a goal position. When path planning is done, the user gives uses the 2D Nav Goal in RVIZ to specify the goal position and orientation. This position (x,y,z) and orientation(x,y,z,w) are actually published in move_base/goal topic. The move_base node links together a global and local planner to accomplish its global navigation task. The Global Planner It is in charge of calculating the safe path in order to arrive at goal pose. Note that the path is calculated before the robot starts moving so it will not take into account the laser readings while moving i.e. unscanned obstacles that are not in the map. Navfn Planner The Navfn planner is the most commonly used global planner for ROS Navigation. It uses Dijkstra's algorithm in order to calculate the shortest path between the initial pose and the goal pose. Carrot Planner This plann...
Read more

Three Wheeled Omnidirectional Robot : Motion Analysis

Image
A Three-wheeled omnidirectional robot is capable to moving in any given direction without rotating by giving velocities to the three wheels in a certain ratio. All motors are situated at an angle of 120 degrees from each other. f B ,  f L ,  f R  are for back , left and right wheel respectively. a x  and a y  are accelerations in X R  and Y R  axis in Robot frame and w is the angular velocity in Robot frame. Hence putting the values of alpha and making it into a matrix form we get : As we need to obtain the f values, we will invert the matrix. We need  a x  , a y  and w is the world frame, so we apply transformation. The transformation angle is    α 1  = 30 degrees.  Putting this in the previous equation, Multiplying the two matrices, we get the final relation as : For example, if you want to move the robot along X axis(world frame), put  a x W  = 1 , ...
Read more

Overview of ATmega328P

Image
While reading about the ATmega328, I came a across a lot of stuff which I would like to share here. There is a humongous amount of information available in its datasheet. Introduction ATmega328 is a single chip micro-controller created by Atmel. It has a In-System Self-programmable Flash memory of 32 KB, SRAM of 2 KB (2048 bytes) and EEPROM of 1 KB (1024 bytes). It has a 8-bit AVR CPU type which means they can process data in 8-bit chunk as its data bus is 8-bit wide.   It contains 23 general purpose I/O pins and 32 general purpose working registers. It has 32 single-byte registers and 8-bit RISC (reduced instruction set computer) devices and hence the name 328. The P is for ' picoPower '. It has 3 flexible timer/counter with compare modes, internal and external interupts , 6-channel 10-bit A/D converter , programmable watch dog timer with internal oscillator and 5 software selectable power saving modes. Digital communication peripherals include 1-USART, 2-SPI and 1-...
Read more