Course Outline

? Day 1: Deep Learning Foundations and Setup

Session 1: Introduction to Deep Learning

• What is deep learning? Key concepts (neurons, layers, activation functions)
• Differences between traditional ML and DL
• Overview of neural network architectures: MLP, CNN, RNN, Transformer

Session 2: Getting Started with TensorFlow and PyTorch

• Installation and environment setup (Google Colab, Jupyter, local)
• TensorFlow vs. PyTorch: Syntax and design philosophy
• Defining tensors, gradients, and computational graphs

Session 3: Building Your First Neural Network

• Feedforward neural network for MNIST digit classification
• Step-by-step model building in TensorFlow and PyTorch
• Loss functions, optimizers, training loops

Lab Activities:

• Tensor basics and operations in both frameworks
• Build and train a simple neural network on MNIST (TF & PyTorch)
• Compare code patterns and training logs

? Day 2: Convolutional Networks and Model Optimization

Session 1: Deep Learning for Computer Vision

• Understanding CNNs: filters, strides, pooling
• Image classification with CIFAR-10 or Fashion-MNIST
• Data augmentation and preprocessing

Session 2: Model Evaluation and Regularization

• Evaluation metrics: accuracy, precision, recall, confusion matrix
• Techniques to avoid overfitting: dropout, early stopping, weight decay
• Visualization of training progress (TensorBoard, matplotlib)

Session 3: Transfer Learning and Pre-trained Models

• Concept of transfer learning and fine-tuning
• Using pretrained models like ResNet, MobileNet (TensorFlow Hub & torchvision)
• Feature extraction vs. full model fine-tuning

Lab Activities:

• Train CNNs on image datasets with both frameworks
• Apply dropout, batch normalization, and tuning techniques
• Fine-tune a pre-trained model and compare accuracy

? Day 3: NLP, Deployment, and Project Showcase

Session 1: Deep Learning for NLP

• Text preprocessing and tokenization
• Text classification using Embeddings and RNNs
• Intro to Transformers and Hugging Face integration

Session 2: Saving, Loading, and Deploying Models

• Saving and loading models (SavedModel, TorchScript)
• Exporting for mobile/web/production
• Serving models using TensorFlow Serving or TorchServe

Session 3: Capstone Project + Wrap-Up

• Mini-project (choose: image or text classification task)
• Build, train, evaluate, and deploy a deep learning model
• Group presentations and discussion on deployment strategies

Lab Activities:

• Train an LSTM or Transformer-based text classifier
• Save and load models in both frameworks
• Deploy a model with Streamlit or Flask demo