Beginner to Advanced with Python & PyTorch

• Course Overview and Learning Objectives
• Introduction to AI and Neural Networks
• What You Will Learn and Build
• Account and Environment Setup (Jupyter, Colab, PyTorch, GitHub)
• Software/Tools Required
• Git Fundamentals: Version Control from Day 1
• Setting up Python and Notebooks

• Python Basics Refresher
• Data Types, Loops, Conditionals
• Functions and Basic I/O
• Six JARs Framework and Expert Systems
• Rule-Based Systems and Decision Making
• Vectors and Matrices for ML
• Numpy Basics, Vector/Matrix Operations
• Hands-on: Linear Algebra with Python

• Understanding Data Types: Numerical, Categorical, Time-Series, Image
• Pandas and Numpy for Preprocessing
• Tokenization and Vectorization Techniques
• Handling Missing Values and Outliers
• Visualizing Data Distributions
• Time-Series Slicing and Windowing
• Image Preprocessing and Augmentation (OpenCV, torchvision)
• Hands-on: Data Pipelines for Deep Learning

• McCulloch-Pitts (MP) Neuron Model
• Binary Inputs/Outputs, Threshold Logic
• The Perceptron Algorithm
• Single-layer Learning
• Python: MP & Perceptron Implementation
• Hands-on: Build and Test in Python

• Applying Theory in Practice
• Project 1.1: Mobile Phone Like/Dislike Classifier

Binary Classification with User Data

• Sigmoid Activation and Learning
• Real-Valued Outputs
• Loss Functions (MSE, Cross-Entropy Intro)
• Gradient Descent: Intuition and Steps
• Python: Sigmoid & Gradient Descent Implementation
• Visualize Learning with Code
• Improve and Tune Gradient Descent

• Probability Essentials for ML
• Bayes’ Theorem, Conditional Probability
• Entropy, KL Divergence, Cross-Entropy
• Sigmoid + Cross Entropy Combined
• Python: Implementing Probabilistic Concepts

• Error Analysis and Improvements
• Project 1.2: Binary Text Classifier
• Project 1.3: Enhanced Text Classifier with Vectorization and Evaluation

• Function Approximation and Universality
• Feedforward Neural Networks (FNNs)
• Layer Types and Activations
• Python: FNN from Scratch
• Training Deep Networks Step-by-step

• Backpropagation: Intuition and Chain Rule
• Manual Derivatives and Scalar Implementation
• Vectorized Backprop for Efficiency
• Python: Full Training Implementation
• Common Pitfalls in Backprop (Vanishing, Exploding Gradients)
• PyTorch Debugging Tools and Gradient Checks

• SGD, Momentum, RMSProp
• Adam, AdaGrad – Pros and Cons
• Custom Optimizer Implementations
• Adaptive Learning Rates and Schedulers
• GPU Acceleration: torch.cuda Essentials

• Project 1.3 Deep Dive:
  • Evaluation Metrics: ROC-AUC, Confusion Matrix, Precision/Recall
  • Using PyTorch’s Model.eval() and torchmetrics
  • Git Branching and Collaborative Coding

• Activation Functions and Initialization (ReLU, Leaky ReLU, Xavier, He)
• Regularization Techniques: L1, L2, Dropout, Early Stopping
• Python: Effects of Initialization and Regularization

• PyTorch Basics: Tensors, Datasets, Modules
• Writing Models and Custom Data loaders
• Training Loops and Model Saving
• Model Check pointing and Resume Training

• Convolution Concepts: Filters, Padding, Strides
• CNN Architectures: LeNet, AlexNet, VGG
• PyTorch: Building CNNs
• Advanced CNNs: ResNet, Inception
• Custom Architectures
• Visualizing Feature Maps and Layers (Grad-CAM)

• Batch Normalization and Dropout
• PyTorch: Implementing Robust CNNs
• Quantization and Knowledge Distillation Overview
• Pruning Networks for Deployment

• Hyperparameter Tuning: Grid Search, Random Search
• Logging and Experiment Tracking using MLflow
• Debugging Training: Visualizing Loss, Gradients, Weights
• Visual Tools: TensorBoard and Custom Dashboards
• Practice: CNN vs FNN on the Same Dataset

• Sequential Data and Applications
• RNNs: Basics and Memory
• Vanishing/Exploding Gradients in RNNs
• LSTM & GRUs: Intuition and Structure
• PyTorch: Implementing LSTM Models
• Real-life Examples: Text, IoT, and Time-Series

• Encoder-Decoder Architectures
• Attention Mechanisms & Transformers (Overview)
• Object Detection: YOLO, SSD, R-CNN
• Model Exporting with TorchScript
• Model Deployment 101:
  • Build an API with Flask/FastAPI
  • Introduction to Docker for ML
  • Model Monitoring and Logging
  • Versioning Models and Data with DVC
  • Deploying on Google AI Platform (Conceptual)

• End-to-End Deep Learning Pipeline
  • Data Preparation
  • Model Design and Training
  • Evaluation and Optimization
  • Deployment and API Creation
  • Git Integration and Collaboration
  • Monitoring and Final Reporting