Neural Network Architectures

In this section, we explore how neural networks are designed and how they truly work, examining both the mathematical principles and the intuitive ideas that drive some of the most widely used neural network architectures.

Topics

• Convolutional Neural Networks
  Explore how CNNs use convolutional layers to automatically extract features from images, enabling powerful solutions for tasks like image classification and object detection.

  • Backpropagtion Intuition Dive in details how backward propagation works in CNN.

• Residual Networks
  Discover how ResNets use skip connections to enable the training of very deep networks, addressing the vanishing gradient problem and improving performance in complex tasks.

• Recurrent Neural Networks
  Learn how RNNs process sequential data by maintaining memory of previous inputs.

• Simple RNN Math Simple math of how a vanilla RNN network is designed.

• LSTM network We dive into the math and intuition of how an LSTM network works.

• Sequence to Sequence Models
  Understand how seq2seq models are designed tohandle tasks like machine translation and text summarization.

• GPT: Stripping the Encoder Away Explore the architecture of the GPT model, adecoder-only transformer model, and understand how it predicts the next tokenin a sequence using self-supervised learning.

• Bert: Bidirectional Encoder Representations from Transformers Learn about the BERT model, a bidirectional transformer architecture that excels in understanding the context of words in a sentence.