Generative AI Course in Chandigarh

Duration: 4 weeks (1 month)

• Objective: Equip learners with theoretical knowledge and practical skills
in generative AI, including foundational models, large language models
(LLMs), LangChain, and real-world applications.
• Prerequisites: Basic knowledge of Python, machine learning concepts, linear algebra, and probability.
• Tools: Python, TensorFlow/PyTorch, Jupyter Notebooks, Hugging Face, LangChain,
NumPy, Pandas, Google Colab.
• Assessment: Weekly assignments and a final capstone project.

Week 1: Introduction and Variational Autoencoders (VAEs)

• Topics:
– Overview of generative AI: Applications (e.g., image generation, text
generation, music synthesis) and types of models (VAEs, GANs, diffusion models, LLMs).
– Key concepts: Probability distributions, likelihood, latent spaces.
– VAEs: Encoder-decoder framework, KL-divergence, reparameterization trick, and applications (e.g., image denoising, data generation).
• Activities:
– Lecture: Introduction to generative AI, its impact, and VAE theory.
– Hands-on: Set up Python environment (TensorFlow/PyTorch, Google
Colab, LangChain) and implement a simple VAE for MNIST digit generation.
– Assignment: Explore a generative AI tool (e.g., DALL·E mini) and train
a VAE on a dataset like Fashion-MNIST, analyzing the latent space.
• Learning Outcomes:
– Understand the scope of generative AI and VAE theory.
– Set up a development environment and generate synthetic data using
VAEs.

• Machine learning and deep learning algorithm
• Mathematics: linear algebra and calculus

• Supervised and unsupervised learning
• libraries like Scikit-Learn
• Algorithms like decision trees , random forests etc.

• Understanding neural networks
• Activation functions, and gradient descent
• Tensor Flow Library
• ANN(Artificial neural network)
• Generative Adversarial Networks (GANs)
• CNN(Convolutional neural networks)
• Variational Autoencoders (VAEs)
• Recurrent Neural Networks (RNNs)

• Generative AI Applications
  
• Understanding Probability and Statistics in Generative AI
  
• Introduction to Generative Models
  
• Deep Learning for Generative Models
  
• Introduction to Generative Adversarial Networks (GANs)
  Autoencoders
  
• Transformers and Attention Mechanisms – “Attention is all you need”.

• Introduction & Applications

•  Probability, Statistics
• Linear Algebra
• Calculus
  
  

• Probabilistic Models (GMMs, HMMs)
  
  

•  (PixelRNN, PixelCNN, Language Models)
• VAEs (Latent space, Reparameterization)
• GANs (Vanilla GAN, cGAN, CycleGAN, StyleGAN, BigGAN)

• (Normalizing Flows, RealNVP, Glow)

• (DDPMs, Stable Diffusion, Imagen, DALL·E)

• Data Processing, Tokenization
•  Pre-training vs Fine-tuning

• zero-shot
• few-shot
• chain-of-thought

• LoRA
• quantization
• distillation

Perplexity

BLEU,

ROUGE

METEOR

• FID
• Inception Score

• PESQ
• MOS

• FVD,
• SSIM