Learn about generative adversarial networks (GANs), a type of AI system where two neural networks compete to create realistic synthetic data. Covered in the IBM AI Fundamentals certification exam.
Table of Contents
Question
You are working on a project that has two convolutional neural networks (CNNs) working against each other. One CNN (the generator) creates new data, while the other CNN (the discriminator) tries to identify if the data is real or generated. This competitive process helps the generator create increasingly realistic outputs.
Which of the following best describes what you are working on?
A. Natural language processing (NLP) system
B. Variational Autoencoder (VAE) Transformer
C. Generative adversarial network (GAN)
D. Convolutional neural network (CNN)
Answer
C. Generative adversarial network (GAN)
Explanation
A GAN is an example of unsupervised learning; it effectively supervises itself and it consists of the generator CNN trying to fool the discriminator CNN. Over time, the generator CNN becomes good enough to fool the discriminator CNN consistently.
The scenario described in the question is a generative adversarial network (GAN).
A GAN consists of two neural networks, a generator and a discriminator, that compete against each other in a zero-sum game:
- The generator network takes random noise as input and tries to generate synthetic data (e.g. images) that look realistic and fool the discriminator. Its goal is to maximize the probability that the discriminator incorrectly classifies the fake data as real.
- The discriminator network evaluates data and tries to distinguish between the real data and the fake data created by the generator. Its goal is to correctly identify which data is real vs fake.
Through many iterations of this adversarial process, the generator learns to create increasingly realistic fake data, while the discriminator gets better at spotting the subtle differences between real and generated data. The result is a generator capable of producing highly realistic synthetic data.
While both networks in this scenario use convolutional layers, making them convolutional neural networks (CNNs), the overall architecture is a GAN. The other options do not accurately describe the competitive two-network structure:
- NLP systems focus on processing and understanding text data rather than generating new data
- Variational Autoencoders (VAEs) have an encoder and decoder but are not adversarial
- CNNs on their own are used for analyzing grid-like data, not generating new data adversarially
So in summary, this project implements a generative adversarial network (GAN), a powerful approach for generating realistic synthetic data. The correct answer is C.
IBM Artificial Intelligence Fundamentals certification exam practice question and answer (Q&A) dump with detail explanation and reference available free, helpful to pass the Artificial Intelligence Fundamentals graded quizzes and final assessments, earn IBM Artificial Intelligence Fundamentals digital credential and badge.