2. Abstract
● Generative Adversarial Networks (GANs) are a type of deep learning algorithm that
can generate new data that is similar to a given training dataset.
● GANs consist of two neural networks: a generator network that produces new data
samples, and a discriminator network that tries to distinguish between the
generated samples and the real ones. The two networks are trained together in an
adversarial setting, where the generator aims to produce samples that are realistic
enough to fool the discriminator, while the discriminator tries to accurately identify
which samples are real and which ones are generated.
3. Outline
● Introduction
● Machine Learning
○ Supervised Learning
○ Unsupervised Learning
● What are GANs?
○ Generative model
○ Discriminative model
● How do GANs work?
● Examples of GANs
● GANs Applications
● Advantages & Disadvantages
● Conclusion
4. Introduction
● Generative Adversarial Networks, or GANs, are a type of neural network architecture that
have gained a lot of attention in recent years due to their ability to generate new, high-
quality, realistic data. GANs were first introduced in 2014 by Ian Goodfellow and his
colleagues.
● The basic idea behind GANs is that they consist of two neural networks: a generator and a
discriminator. The generator network takes random noise as input and produces a new
piece of data, such as an image or a sound clip. The discriminator network takes this new
piece of data, along with real examples of the same type of data, and tries to distinguish
between them. The two networks are trained simultaneously, with the generator trying to
produce data that is indistinguishable from the real data, and the discriminator trying to
correctly identify which data is real and which is generated.
● GANs have a wide range of applications, including generating realistic images, videos, and
music, as well as in natural language processing tasks such as text generation and machine
translation.
5. Machine Learning
Machine learning is a type of artificial intelligence (AI) that allows computers to learn
and improve from experience without being explicitly programmed. In other words,
instead of being told what to do, machines can learn from data and make
predictions or decisions based on that data.
Types of machine learning:
● Supervised Learning
● Unsupervised Learning
● Reinforcement Learning
6. Machine Learning
● Supervised learning is a type of
machine learning that uses labeled
data to train machine learning models.
In labeled data, the output is already
known. The model just needs to map
the inputs to the respective outputs.
● This correction of the model is generally
referred to as a supervised form of
learning, or supervised learning.
● Examples: Linear Regression, Logistic
Regression, Decision Tree.
Example of Supervised Learning
7. Machine Learning
● Unsupervised learning is a type of machine learning
that uses unlabeled data to train machines. Unlabeled
data doesn’t have a fixed output variable. The model
learns from the data, discovers the patterns and
features in the data, and returns the output.
● Examples: problems include clustering, generative
modeling, learning algorithms are K-means and
Generative Adversarial Networks. Example of Unsupervised Learning
8. What are GANs?
● Generative adversarial networks are implicit likelihood models that generate
data samples from the statistical distribution of the data. They’re used to copy
variations within the dataset. They use a combination of two networks:
generator and discriminator.
9. What are GANs?
● The generator generates a batch of
samples, and these, along with real
examples from the domain, are
provided to the discriminator and
classified as real or fake.
● The discriminator is then updated to
get better at discriminating real and
fake samples in the next round, and
importantly, the generator is updated
based on how well, or not, the
generated samples fooled the
discriminator.
10. What are GANs?
● They both play an adversarial game where the generator tries to fool the
discriminator by generating data similar to those in the training set. The
Discriminator tries not to be fooled by identifying fake data from real data.
17. GANs Applications
● With the help of DCGANs, you can train images of cartoon characters for
generating faces of anime characters as well as Pokemon characters.
18. GANs Applications
● GANs can be trained on the images of humans to generate realistic
faces. The faces that you see below have been generated using GANs
and do not exist in reality.
19. Advantages
● GANs can generate high-quality, realistic synthetic data that closely
resembles real-world data.
● GANs have a wide range of applications, including image and video
synthesis, text-to-image generation, and music synthesis, among others.
● GANs can be used to generate data with specific characteristics, such as
generating images of a particular style or genre.
● GANs can be used for data augmentation, which can help improve the
performance of machine learning models.
20. Disadvantages
● GANs can be challenging to train, and the training process can be
unstable and time-consuming.
● GANs can suffer from mode collapse, where the generator produces a
limited set of outputs that do not represent the full range of possible
outputs.
● GANs can produce biased outputs if the training data is biased, which
can perpetuate and amplify existing biases in the data.
● GANs can be computationally expensive, particularly for large and
complex datasets.
21. Conclusion
GANs have a wide range of applications, including image and video
synthesis, text-to-image generation, and music synthesis, among others.
However, GANs also face several challenges, including instability during
training, mode collapse, and difficulty in controlling the output.
Therefore, researchers are continually developing new variants of GANs
and exploring techniques to address these challenges.