Machine Learning and Its Engineering Applications Course

Lesson 1 • What Machine Learning Is

Defines ML, distinguishes it from traditional programming, and maps its relationship to AI and data science. Establishes vocabulary used throughout the course.

Lesson 2 • Core Learning Paradigms

Contrasts supervised, unsupervised, and reinforcement learning with concrete examples. Enables students to match problem types to the correct learning paradigm.

Lesson 3 • The ML Project Lifecycle

Outlines the end-to-end workflow from problem framing to model deployment and monitoring. Sets expectations for the full engineering scope covered in subsequent chapters.

Lesson 4 • Essential Mathematics for ML

Reviews linear algebra, calculus, probability, and statistics concepts that underpin ML algorithms. Provides the quantitative fluency needed for later algorithm study.

Lesson 1 • Exploratory Data Analysis

Applies statistical summaries and visualizations to uncover distributions, outliers, and relationships. Informs preprocessing and feature engineering strategies in later sections.

Lesson 2 • Data Cleaning and Imputation

Addresses missing values, duplicates, inconsistent formats, and noise in raw datasets. Ensures training data integrity before model fitting.

Lesson 3 • Building Reproducible Data Pipelines

Designs automated, version-controlled pipelines that apply transformations consistently across splits. Supports reliable experimentation and production deployment.

Lesson 4 • Data Collection and Sourcing

Examines methods for gathering data from databases, APIs, web scraping, and sensors. Grounds data sourcing decisions in quality and representativeness requirements.

Lesson 1 • Model Evaluation and Selection

Applies cross-validation, performance metrics, and statistical tests to compare models objectively. Prevents overfitting and guides final model selection decisions.

Lesson 2 • Linear and Logistic Regression

Derives linear regression from first principles and extends it to logistic regression for classification. Establishes the loss-function and gradient-descent foundation used by all later models.

Lesson 3 • Decision Trees and Rule-Based Models

Explains tree splitting criteria, pruning, and interpretability advantages. Serves as the building block for ensemble methods introduced next.

Lesson 4 • Ensemble Methods

Covers bagging, boosting, and stacking to reduce variance and bias beyond single-model limits. Equips students with the most competitive off-the-shelf algorithms for tabular data.

Lesson 1 • Anomaly and Outlier Detection

Introduces statistical, distance-based, and model-based anomaly detection for quality control and fault detection. Directly applicable to manufacturing and infrastructure monitoring.

Lesson 2 • Dimensionality Reduction Techniques

Applies PCA, t-SNE, and UMAP to reduce feature space while preserving structure. Improves visualisation, training speed, and noise reduction in high-dimensional datasets.

Lesson 3 • Association Rule Learning

Mines frequent itemsets and association rules using Apriori and FP-Growth algorithms. Supports recommendation, maintenance bundling, and process optimisation use cases.

Lesson 1 • Hyperparameter Tuning and Architecture Search

Applies grid search, random search, and Bayesian optimisation to find optimal network configurations. Reduces trial-and-error and accelerates model development cycles.

Lesson 2 • Feedforward Neural Network Fundamentals

Derives the perceptron, multilayer architecture, and backpropagation algorithm step by step. Establishes the computational graph intuition used in all deep learning frameworks.

Lesson 3 • Recurrent Networks and Sequence Models

Covers RNNs, LSTMs, and GRUs for modelling temporal dependencies in sequential engineering data. Prepares students for time-series forecasting and process control applications.

Lesson 4 • Training Deep Networks Effectively

Addresses optimisation algorithms, regularisation, and normalisation techniques that stabilise deep network training. Prevents common failure modes such as vanishing gradients and overfitting.

Lesson 1 • MLOps and CI/CD for ML

Integrates ML workflows into continuous integration and delivery pipelines with automated testing and retraining triggers. Ensures reproducibility and rapid iteration in production.

Lesson 2 • Scalable Training Infrastructure

Covers distributed training strategies, GPU utilisation, and cloud compute orchestration for large-scale models. Reduces training time and enables experimentation at scale.

Lesson 3 • Model Serving and Deployment Patterns

Compares batch inference, real-time REST APIs, and edge deployment architectures. Matches deployment pattern to latency, throughput, and resource constraints.

Lesson 4 • Monitoring and Observability

Implements data drift, concept drift, and performance degradation detection in deployed models. Enables proactive maintenance before model failures affect engineering operations.

Lesson 1 • Energy Systems and Demand Forecasting

Forecasts energy demand and optimises grid operations using time-series and optimisation models. Supports renewable integration and load balancing in smart energy systems.

Lesson 2 • Process Optimisation and Control

Uses regression and reinforcement learning to optimise manufacturing process parameters and control loops. Improves yield, energy efficiency, and throughput in production systems.

Lesson 3 • Predictive Maintenance Systems

Builds failure prediction models from sensor time-series using survival analysis and classification. Reduces unplanned downtime and maintenance costs in industrial settings.

Lesson 4 • Structural Health Monitoring

Detects damage and degradation in civil and mechanical structures using vibration and acoustic sensor data. Enables condition-based maintenance for bridges, pipelines, and machinery.

Lesson 1 • Adversarial Robustness and Security

Examines adversarial attacks, data poisoning, and model theft, then applies defences. Protects ML systems deployed in critical infrastructure and safety-sensitive environments.

Lesson 2 • Fairness and Bias Mitigation

Identifies sources of bias in training data and model outputs, then applies mitigation techniques. Ensures equitable outcomes in workforce, resource allocation, and safety systems.

Lesson 3 • Uncertainty Quantification

Quantifies epistemic and aleatoric uncertainty using Bayesian methods and conformal prediction. Enables engineers to act on model confidence levels in safety-critical decisions.

Lesson 4 • Model Explainability and Interpretability

Applies SHAP, LIME, and attention visualisation to explain model predictions to engineers and stakeholders. Builds trust and supports regulatory compliance in high-stakes applications.