Machine Learning and Deep Learning Course

Lesson 1 • Data Representation and Preprocessing

Teaches how raw data is transformed into model-ready features. Directly enables effective model training in later chapters.

Lesson 2 • Python Toolkit for ML

Introduces NumPy, Pandas, and Matplotlib as the core data manipulation and visualisation stack. Prepares students for hands-on coding in all subsequent chapters.

Lesson 3 • What Machine Learning Is

Defines ML, its relationship to AI and statistics, and the types of learning. Anchors all subsequent technical content in a unified conceptual map.

Lesson 4 • Essential Mathematics for ML

Covers linear algebra, calculus, probability, and statistics needed for ML algorithms. Provides the quantitative language used throughout the course.

Lesson 1 • Model Evaluation and Selection

Provides rigorous techniques for measuring and comparing model performance. Ensures students can make defensible model-selection decisions.

Lesson 2 • Hyperparameter Tuning

Covers systematic search strategies for optimising model configuration. Bridges the gap between a working model and a production-ready one.

Lesson 3 • Decision Trees and Ensemble Methods

Explains tree-based splitting criteria and extends them to powerful ensemble models. Students gain tools for high-accuracy tabular data modelling.

Lesson 4 • Linear and Logistic Regression

Covers ordinary least squares regression and logistic regression for binary classification. Establishes the loss-minimisation paradigm central to all ML models.

Lesson 1 • Dimensionality Reduction

Teaches PCA, t-SNE, and UMAP for compressing high-dimensional data. Enables visualisation and improved downstream model performance.

Lesson 2 • Anomaly Detection Methods

Covers statistical and model-based approaches to identifying outliers. Directly applicable to fraud detection, quality control, and monitoring systems.

Lesson 3 • Clustering Algorithms

Introduces partitional and hierarchical clustering methods for grouping similar data points. Builds intuition for distance metrics and cluster quality evaluation.

Lesson 1 • Forward and Backward Propagation

Derives the mathematics of forward pass computation and gradient backpropagation. Students can manually trace data flow and gradient updates through a network.

Lesson 2 • Biological to Artificial Neurons

Maps biological neuron structure to the perceptron model and multilayer networks. Grounds deep learning architecture in an intuitive conceptual foundation.

Lesson 3 • Optimisation Algorithms

Compares gradient descent variants and adaptive optimisers for training stability. Directly impacts convergence speed and final model accuracy.

Lesson 4 • Regularisation in Neural Networks

Teaches dropout, batch normalisation, and weight decay to combat overfitting. Prepares students to build networks that generalise to unseen data.

Lesson 1 • CNN Training Best Practices

Covers data augmentation pipelines, mixed-precision training, and debugging strategies. Ensures students can efficiently train CNNs on real datasets.

Lesson 2 • Classic CNN Architectures

Surveys LeNet, AlexNet, VGG, ResNet, and EfficientNet design principles. Teaches architectural evolution and the reasoning behind each design choice.

Lesson 3 • Object Detection and Segmentation

Extends classification CNNs to localisation and pixel-level labelling tasks. Covers YOLO, Faster R-CNN, and U-Net for real-world vision applications.

Lesson 4 • Convolution and Pooling Operations

Explains the convolution operation, filter learning, and spatial downsampling. Establishes the core computational primitives of all CNN architectures.

Lesson 1 • LSTM and GRU Architectures

Details gating mechanisms that enable long-range dependency learning. Students implement and compare LSTM and GRU on benchmark sequence tasks.

Lesson 2 • Sequence-to-Sequence Models

Covers encoder-decoder architectures for variable-length input-output mapping. Directly applicable to machine translation, summarisation, and chatbots.

Lesson 3 • Time-Series Forecasting with RNNs

Applies recurrent models to univariate and multivariate forecasting problems. Connects sequence modelling theory to business and scientific forecasting use cases.

Lesson 1 • Prompt Engineering and In-Context Learning

Teaches zero-shot, few-shot, and chain-of-thought prompting strategies. Enables students to extract maximum performance from LLMs without retraining.

Lesson 2 • Fine-Tuning Pretrained Language Models

Applies Hugging Face Transformers to classification, NER, and QA tasks. Gives students a practical workflow for adapting LLMs to domain-specific problems.

Lesson 3 • Parameter-Efficient Fine-Tuning

Covers LoRA, prefix tuning, and prompt tuning for low-resource adaptation. Reduces compute and memory costs while maintaining competitive performance.

Lesson 4 • Self-Attention and the Transformer

Derives scaled dot-product attention and multi-head attention from first principles. Explains why Transformers replaced RNNs as the dominant sequence architecture.

Lesson 1 • Experiment Tracking and Reproducibility

Uses MLflow and Weights & Biases to log experiments and manage artefacts. Establishes the discipline needed for collaborative, reproducible ML research.

Lesson 2 • ML Pipelines and Workflow Orchestration

Builds automated training pipelines using tools like Airflow and Kubeflow. Enables continuous retraining and reduces manual intervention in production.

Lesson 3 • Monitoring and Data Drift Detection

Implements statistical tests and dashboards to detect model and data drift. Ensures sustained model performance after deployment in changing environments.

Lesson 4 • Model Serving and APIs

Deploys models as REST APIs using FastAPI and containerised microservices. Connects trained models to real-world applications and end users.