A Generative Journey to AI: Mastering the foundations and frontiers of generative deep learning/ Генеративное путешествие к ИИ: изучаем основы и границы генеративного глубокого обучения
Год издания: 2025
Автор: Ramchandani Toni / Рамчандани Тони
Издательство: BPB Publications
ISBN: 978-93-65890-846
Язык: Английский
Формат: PDF/EPUB
Качество: Издательский макет или текст (eBook)
Интерактивное оглавление: Да
Количество страниц: 553
Описание: Explore the world of generative AI, a technology capable of creating new data that closely resembles reality. This book covers the fundamentals and advances through cutting-edge techniques. It also clarifies complex concepts, guiding you through the essentials of Deep Learning, neural networks, and the exciting world of generative models, like GANs, VAEs, Transformers, etc.
This book introduces Deep Learning, Machine Learning, and neural networks as the foundation of generative models, covering types like GANs and VAEs, diffusion models, and other advanced architectures. It explains their structure, training methods, and applications across various fields. It discusses ethical considerations, responsible development, and future trends in generative AI. It concludes by highlighting how generative AI can be used creatively, transforming fields like art and pushing the boundaries of human creativity, while also addressing the challenges of using these technologies responsibly.
This book provides the tools and knowledge needed to leverage generative AI in real-world applications. By the time you complete it, you will have a solid foundation and the confidence to explore the frontiers of AI.
Key Features:
- Comprehensive guide to mastering generative AI and Deep Learning basics.
- Covers text, audio, and video generation with practical examples.
- Insights into emerging trends and potential advancements in the field.
What you will learn:
- Understand the fundamentals of deep learning and neural networks.
- Master generative models like GANs, VAEs, and Transformers.
- Implement AI techniques for text, audio, and video creation.
- Apply generative AI in real-world scenarios and applications.
- Navigate ethical challenges and explore the future of AI.
Who this book is for:
This book is ideal for AI enthusiasts, developers, and professionals with a basic understanding of Python programming and Machine Learning.
Исследуйте мир генеративного ИИ - технологии, способной создавать новые данные, максимально приближенные к реальности. В этой книге рассматриваются основы и достижения, достигнутые с помощью передовых технологий. Она также разъясняет сложные концепции, знакомя вас с основами глубокого обучения, нейронными сетями и захватывающим миром генеративных моделей, таких как GAN, VAE, трансформеры и т.д.
Эта книга знакомит с глубоким обучением, машинным обучением и нейронными сетями как основой генерирующих моделей, охватывая такие типы, как GAN и VAE, диффузионные модели и другие продвинутые архитектуры. В ней объясняется их структура, методы обучения и приложения в различных областях. В ней обсуждаются этические соображения, ответственное развитие и будущие тенденции в области генеративного искусственного интеллекта. В заключение в ней рассказывается о том, как можно творчески использовать генеративный ИИ, преобразуя такие области, как искусство, и расширяя границы человеческого творчества, а также решая проблемы ответственного использования этих технологий.
В этой книге представлены инструменты и знания, необходимые для использования генеративного ИИ в реальных приложениях. К тому времени, когда вы завершите его, у вас будет прочная основа и уверенность в том, что вы сможете исследовать границы искусственного интеллекта.
Ключевые функции:
- Исчерпывающее руководство по освоению генеративного искусственного интеллекта и основ глубокого обучения.
- Создание текстов, аудио и видео с практическими примерами.
- Понимание новых тенденций и потенциальных достижений в этой области.
Чему вы научитесь:
- Понимание основ глубокого обучения и нейронных сетей.
- Освоение генеративных моделей, таких как GAN, VAE и трансформеры.
- Внедрять методы искусственного интеллекта для создания текстов, аудио и видео.
- Применять генеративный искусственный интеллект в реальных сценариях и приложениях.
- Решать этические проблемы и изучать будущее искусственного интеллекта.
Для кого предназначена эта книга:
Эта книга идеально подходит для энтузиастов искусственного интеллекта, разработчиков и профессионалов, имеющих базовое представление о программировании на Python и машинном обучении.
Примеры страниц (скриншоты)
Оглавление
1. Introduction to Deep Learning
Introduction
Structure
Objectives
Deep learning from generative AI lens
Understanding deep learning
The evolution of deep learning
Mathematics for deep learning
Calculus in deep learning
Probability theory in deep learning
Machine learning
Types of machine learning
Supervised learning
Unsupervised learning
Reinforcement learning
Deep learning vs. traditional machine learning
Real-world applications of deep learning
Conclusion
2. Neural Networks and Deep Learning Architectures
Introduction
Structure
Objectives
Understanding neurons and layers
Biological neuron inspiration
Activation functions and their importance
Types of activation functions
Activation functions visualization
Forward propagation in neural networks
Importance of feedforward networks
Steps in forward propagation
The concept of backpropagation
Training neural networks
Multilayer Perceptron
Training MLPs
Overfitting
Convolutional neural networks
Shared weights
Shared biases
Recurrent neural networks
SimpleRNN implementation
Autoencoders
Introduction to generative models
The quest for creativity
Bridging the gap with reality
Unleashing imagination
Practical applications
The intricacies of training
Fueling curiosity and exploration
Conclusion
3. Unveiling Generative Models
Introduction
Structure
Objectives
Understanding generative models
Taxonomy of generative models
Selecting the right model
Basic generative model
Discriminative models
Types of discriminative models
Generative models vs. discriminative models
Mathematical notation
Bayes’ theorem in action
Parameter estimation and learning
Combining generative and discriminative models
Applications of generative models
Applications of discriminative models
Latent space
Evaluating generative models
Bias in generative AI models
Conclusion
4. Generative Adversarial Networks
Introduction
Structure
Objectives
Introduction to GANs
Advantages of GANs
Disadvantages of GANs
Variants of GANs
Applications of GANs
Exploring vital concepts
Kullback-Leibler divergence
Jensen-Shannon divergence
Nash equilibrium
Objective functions
Training GANs
GAN stability strategies
Feature matching
Minibatch discrimination
Historical averaging
One-sided label smoothing
Batch normalization
Implementing vanilla GAN
Importing libraries
Loading and pre-processing the dataset
Defining the generator
Defining the discriminator
Compiling the discriminator and GAN Model
Defining a function to have generated images
Training the GAN
Output
Implementing Deep Convolutional GAN
DCGANs working
Importing libraries
Loading and pre-processing data
Data conversion and reshaping
Generator network
Discriminator network
Compiling the discriminator
Compiling the combined DCGAN model
Saving generated images function
Training loop
Output
Conclusion
5. Variational Autoencoders
Introduction
Structure
Objectives
Introduction to variational autoencoders
History of VAEs
Architecture
Understanding the encoder
Understanding the decoder
Mathematically
Evidence lower bound
Reparameterization trick
Variations of VAEs
Applications of VAEs
Limitations of VAEs
Vanilla VAE
Conditional VAE
GANs vs. VAEs
Fusion of GAN-VAE
Challenges and limitations
Conclusion
6. Diffusion Models
Introduction
Structure
Objectives
Introduction to diffusion models
History
Denoising diffusion models
Forward diffusion process
Reverse diffusion process
Markov chain modeling
Neural network parameterization
Annealing techniques
Implementing diffusion model
Variants of diffusion models
Sampling-acceleration enhancement
Likelihood-maximization enhancement
Data-generalization enhancement
Application of diffusion models
Limitations of diffusion models
U-Net
GANs vs. VAEs vs. diffusion models
Diffusion-GANs
Stable Diffusion
Conclusion
7. Transformers and Large Language Models
Introduction
Structure
Objectives
Introducing Transformers
Transformer architecture
Encoder and decoder stack
Attention mechanism
Scaled dot-product attention
Multi-head attention
Positional encoding
Feedforward neural networks
Layer normalization and residual connections
Importance in Transformers
Output and probability prediction with softmax
Training Transformers
Variants of Transformers
Applications of Transformers
LLMs
Hugging Face
BERT
Pre-training BERT
Fine-tuning BERT
BERT variants
Implementing BERT question-answering system
GPT
Implementing GPT text generation
GPT-4
Vision Transformers
Basic concept and architecture
Training and application
T5
Conclusion
8. Exploring Generative Models
Introduction
Structure
Objectives
Pix2Pix
Implementation
Execution
CycleGANs
ProGANs
StyleGAN-XL
Evolution of StyleGAN models
Key features and improvements of StyleGAN-XL
Limitations and challenges of StyleGAN-XL
Autoregressive models
Mathematical foundations of ARMs
Types of autoregressive models
Image generation models
Applications of autoregressive models
Implementing LSTM
Importing libraries
Implementing PixelCNN
Energy-based models
Architectural variants of EBMs
Mathematics underlying energy-based models
Training energy-based models
Langevin dynamics
Applications of EBMs
Challenges and limitations
Normalizing Flow Models
Brief history
Probability density function transformation
Types of flows
Architecture
Types of flow transformations
RealNVP
GLOW
FFJORD
Mathematical foundation of FFJORD
The Gemini Model
Conclusion
9. Video and Music Generation
Introduction
Structure
Objectives
Video generation
Evolution of video generation
Fundamentals of video generation
Types of video generative models
Datasets for video generative models
Applications of video generative models
Video GANs
Strategies in Video GANs
Unconditional video generation
Conditional video generation
Motion and content generative adversarial network
Conditional MoCoGANs
Concept of conditional MoCoGAN
Architecture and mechanism
Training and adversarial process
Diffusion-based video generative model
VideoCrafter1, open diffusion model
Transformer-based video generative model
MAsked Generative VIdeo Transformer
Frame interpolation models
Implementation
Advanced video generation models
Music generation
Evolution of music generation
Fundamentals of music generation
Dataset for music generation
Applications of music generation
MuseGAN
Data representation
Temporal structure modeling
MidiNet
LSTM based music generation system
Implementation
Conclusion
10. Artistic Side of Generative AI
Introduction
Structure
Objectives
DeepDream
Implementation
Neural style transfer
Mathematically
Implementation
Playing games with generative AI
Deep reinforcement learning
Deep Reinforcement Learning in generative AI
Deep Q-learning
Generative Adversarial Imitation Learning
Generating 3D structures from a 2D Slice
Deepfakes
Dynamics of GANs in deepfake generation
StyleGAN, an advanced GAN for realistic image synthesis
Strategies and challenges of deepfake detection
Notable examples of deepfake creation and detection methods
Conclusion
11. Ethics, Challenges, and Future
Introduction
Structure
Objectives
Ethical generative AI
Historical context and evolution of AI ethics
Key ethical concerns
Privacy and data security
Deepfakes, misinformation, and manipulation
Bias and discrimination in AI algorithms
Intellectual property and creative rights
AI in surveillance and implications for human rights
Socioeconomic impacts and the digital divide
Case studies in ethical generative AI
Case study 1: Political deep fakes
Case study 2: Bias in recruitment algorithms
Case study 3: AI in healthcare diagnostics
Case study 4: AI generated art and copyright
Case study 5: AI in financial market predictions
Case study 6: Surveillance AI and privacy
Case study 7: AI in content moderation
Global regulatory framework and legal considerations
National and regional approaches to AI regulation
International guidelines and agreements
Industry standards and self-regulation
Legal challenges in AI regulation
The future of AI regulation
Best practices for ethical AI development
Ensuring transparency and accountability
Ethical AI auditing and certification
Corporate policies and ethical AI governance
Continuous learning and adaptation
Engagement with stakeholders and the public
Examples of the industry
The role of AI ethics in organizational culture
AI hallucinations
Ethical implications of AI hallucinations
Challenges in minimizing hallucinations in AI
Case studies and examples of AI hallucinations
Mitigating hallucinations in AI
Future challenges and opportunities in ethical AI
Responsible generative AI
AI Verify
Road ahead for ethical generative AI
Conclusion
Index
