Automated Machine Learning (AutoML) Course

Lesson 1 • Types of AutoML Systems

Distinguishes full-pipeline AutoML, neural architecture search, and meta-learning systems. Helps students match system type to problem requirements.

Lesson 2 • What AutoML Is and Why It Matters

Defines AutoML as the automation of end-to-end ML workflows and contrasts it with manual model development. Motivates the efficiency and democratization benefits.

Lesson 3 • Core Machine Learning Concepts

Covers supervised, unsupervised, and reinforcement learning paradigms essential for understanding AutoML targets. Establishes vocabulary used throughout the course.

Lesson 4 • Setting Up the AutoML Environment

Guides installation of Python, key libraries, and a local AutoML framework. Students leave with a functional development environment ready for hands-on labs.

Lesson 1 • Understanding Data Quality Issues

Identifies missing values, outliers, duplicates, and schema inconsistencies as primary data quality threats. Connects poor data quality to degraded AutoML performance.

Lesson 2 • Train, Validation, and Test Splitting

Explains holdout, k-fold, and stratified splitting strategies to prevent data leakage. Students configure splits that yield unbiased AutoML evaluations.

Lesson 3 • Handling Missing and Imbalanced Data

Teaches imputation strategies and resampling methods to address gaps and class imbalance. Students apply these techniques before feeding data into AutoML systems.

Lesson 4 • Feature Encoding and Scaling

Covers one-hot encoding, ordinal encoding, and normalization methods required by most AutoML frameworks. Ensures features are numerically compatible with downstream algorithms.

Lesson 1 • Evaluating Feature Engineering Impact

Teaches feature importance scoring and ablation studies to measure engineering value. Students quantify the contribution of automated features to final model performance.

Lesson 2 • Dimensionality Reduction in AutoML

Explains PCA, autoencoders, and manifold methods as automated dimensionality reduction options. Students apply reduction to high-dimensional datasets within an AutoML workflow.

Lesson 3 • Automated Feature Generation Tools

Introduces tools that automatically synthesize new features from raw columns using aggregations and transformations. Students run automated generation on a structured dataset.

Lesson 4 • Principles of Feature Engineering

Reviews manual feature engineering concepts as a baseline for understanding automation. Establishes why feature quality drives model performance more than algorithm choice.

Lesson 1 • Hyperparameters vs. Model Parameters

Clarifies the distinction between learned parameters and user-defined hyperparameters. Establishes why hyperparameter choice critically affects model generalization.

Lesson 2 • Grid Search and Random Search

Implements exhaustive grid search and stochastic random search as baseline HPO methods. Students benchmark both approaches on a classification task.

Lesson 3 • Bayesian Optimization for HPO

Explains surrogate model-based Bayesian optimization as a sample-efficient HPO strategy. Students configure and run Bayesian search using a popular AutoML library.

Lesson 4 • Multi-Fidelity and Early Stopping Methods

Covers Hyperband and successive halving to allocate compute budgets efficiently. Students reduce HPO runtime without sacrificing final model quality.

Lesson 1 • Automated Algorithm Selection Strategies

Explains combined algorithm selection and hyperparameter optimization (CASH) as the core AutoML search problem. Students run CASH on a benchmark dataset.

Lesson 2 • Algorithm Selection Fundamentals

Reviews the algorithm landscape—linear models, trees, SVMs, and neural nets—as candidates for automated selection. Connects algorithm properties to dataset characteristics.

Lesson 3 • AutoML Ensemble Post-Processing

Demonstrates automated ensemble selection and post-hoc pruning to balance accuracy and complexity. Students finalize an ensemble and prepare it for deployment.

Lesson 4 • Bagging and Boosting Ensembles

Covers bagging variance reduction and boosting sequential error correction as ensemble foundations. Students implement both within an AutoML pipeline.

Lesson 1 • NAS Search Strategies

Compares reinforcement learning, evolutionary, and gradient-based NAS search strategies. Students select an appropriate strategy based on compute budget and task type.

Lesson 2 • Efficient NAS with Proxies and Surrogates

Introduces performance predictors and low-fidelity proxies to reduce NAS compute cost. Students apply proxy-based evaluation to speed up architecture search.

Lesson 3 • NAS for Tabular and Time-Series Data

Extends NAS beyond vision tasks to tabular and sequential data architectures. Students run a NAS experiment on a non-image dataset and analyze results.

Lesson 4 • Evaluating and Deploying NAS Results

Covers architecture evaluation protocols, retraining from scratch, and exporting NAS-discovered models. Students produce a deployment-ready NAS model.

Lesson 1 • Multi-Modal AutoML Pipelines

Combines text, image, and tabular modalities in a single automated pipeline. Students design a fusion strategy and evaluate multi-modal model performance.

Lesson 2 • AutoML for Computer Vision

Applies NAS and transfer learning automation to image classification and object detection. Students run an AutoML vision pipeline on a labeled image dataset.

Lesson 3 • AutoML for Time-Series Forecasting

Addresses automated feature extraction, model selection, and horizon tuning for time-series data. Students forecast a real-world time series using an AutoML framework.

Lesson 4 • AutoML for Natural Language Processing

Covers automated text preprocessing, embedding selection, and fine-tuning of language models. Students build an AutoML NLP pipeline for classification and sentiment tasks.

Lesson 1 • MLOps Integration for AutoML

Integrates AutoML runs into CI/CD pipelines, experiment tracking, and model registries. Students automate retraining triggers and version control for AutoML artifacts.

Lesson 2 • Fairness, Bias, and Explainability

Addresses automated bias auditing, fairness constraints, and explainability tools for AutoML outputs. Students generate fairness reports and SHAP explanations for a deployed model.

Lesson 3 • Monitoring Model Performance in Production

Teaches data drift, concept drift, and performance degradation detection for deployed AutoML models. Students configure monitoring dashboards and alerting thresholds.

Lesson 4 • Governance, Compliance, and Documentation

Establishes model cards, audit trails, and regulatory alignment practices for AutoML systems. Students produce a governance-ready model card and deployment checklist.