Advanced
Deep Learning & AI

1Beyond Linear Models

• Limitations of simple regression/classification
• Ensemble Learning: Random Forests
• Gradient Boosting concepts (XGBoost/LightGBM)

2Feature Engineering

• Encoding categorical variables (One-Hot vs Label)
• Normalization and Standardization
• Handling imbalanced datasets

3Advanced Evaluation

• Confusion Matrices
• Precision, Recall, and F1-Score
• Cross-validation techniques

4Hands-on project: Customer Churn Predictor

Build a robust model to predict which customers are likely to cancel a subscription service using Random Forests and analyzing feature importance.

1Introduction to TensorFlow/Keras

• The architecture of a Neural Network
• Input, Hidden, and Output layers
• Activation functions (ReLU, Sigmoid, Softmax)

2Training a Neural Net

• Loss functions and Optimizers (Adam, SGD)
• Epochs and Batch sizing
• Visualizing the training history (Loss/Accuracy curves)

3Hands-on project: Handwritten Digit Recognition

The "Hello World" of Deep Learning. Build a multi-layer network trained on the MNIST dataset to identify handwritten numbers (0-9) with high accuracy.

1Convolutional Neural Networks

• How computers "see" images (Pixel arrays)
• Convolutions and Filters
• MaxPooling and Flattening

2Image Data Processing

• Resizing and normalizing images
• Data Augmentation (rotating, flipping images to increase data)

3Hands-on project: Image Classifier

Build an AI that can distinguish between two distinct image categories (e.g., "Cat vs Dog" or "Pizza vs Burger") using a Convolutional Neural Network.

1Processing Text Data

• Tokenization and removing Stop Words
• Stemming and Lemmatization
• Converting words to numbers (Bag of Words / TF-IDF)

2Sentiment Analysis

• Understanding context in text
• Using pre-trained models vs training your own

3Hands-on project: Spam Message Detector

Create an NLP model that reads SMS or Email text and classifies it as "Spam" or "Ham" (safe) based on keyword patterns and text structure.