Syllabus

#SESSION 1 (LIVE IN-PERSON)

INTRODUCTION

Introduction of the course. The basic concepts of AI and their application in ML will be introduced. Topics will include the general definition of Intelligence, knowledge representation, basic algorithms, and the analysis of typical intractable problems and heuristic approaches. Book Chapters: Machine Learning Engineering (Chapter 1: Introduction)

#SESSION 2 (LIVE IN-PERSON)

THE ANATOMY OF A MACHINE LEARNING PROJECT

Exploration of the general pipeline for ML projects. Definition of project scope, goals, required domain knowledge, and formation of a multidisciplinary team.

Book Chapters: Machine Learning Engineering (Chapter 2: Before the Project Starts)

#SESSION 3 (LIVE IN-PERSON)

THE IMPORTANCE OF DATA. DATA COLLECTION AND PREPARATION

Analysis of typical data issues (missing values, imputation, risks of data leakage, outliers, etc.) Data transformation (hot encoding and other methods).

Book Chapters: Machine Learning Engineering (Chapter 3: Data Collection and Preparation)

#SESSION 4 (LIVE IN-PERSON)

FEATURE ENGINEERING

Assignment I: DATA PREPARATION ASSIGNMENT

The concept of feature. Feature extraction and feature selection (methods available). Complexity analysis and dimensionality reduction.

Book Chapters: Machine Learning Engineering (Chapter 4: Feature Engineering)

#SESSION 5 (LIVE IN-PERSON)

FUNDAMENTAL ALGORITHMS

Taxonomy of main algorithms. Supervised vs Unsupervised. Parametric vs Non-parametric. Instance-based vs Model-based. Manual feature extraction vs Representational methods. Reflex vs state and variable-based models. General principles: Loss functions and gradient descent.

#SESSION 6 (LIVE IN-PERSON)

ANATOMY OF AN ML ALGORITHM. SUPERVISED MODEL TRAINING I

General architecture. Training methods. Task analysis (regression vs classification). Performance metrics. Overfitting and generalization.

Book Chapters: Machine Learning Engineering (Chapter 5: Supervised Model Training (Part 1))

#SESSION 7 (LIVE IN-PERSON)

ANATOMY OF AN ML ALGORITHM. SUPERVISED MODEL TRAINING II

Model training and staking. Troubleshooting and best practices.

Book Chapters: Machine Learning Engineering (Chapter 6: Supervised Model Training (Part 2))

#SESSION 8 (LIVE IN-PERSON)

PRACTICE

Assignment II: END-to-END DEMO

End-to-end practical case. From EDA to model training and validation. Testing of various classical machine learning algorithms.

#SESSION 9 (LIVE IN-PERSON)

COMMON ISSUES

Managing imbalanced classes. Sampling biases. Handling uncertainty (epistemic vs aleatoric).

#SESSION 10 (LIVE IN-PERSON)

ML INTEGRATION. PIPELINES, ENSEMBLE METHODS, HYPER-TUNING

Ensemble methods. Pipelines of algorithms. Methods for hyper-parameter tuning. The future of AutoML. Bayesian optimization of parameters.

#SESSION 11 (LIVE IN-PERSON)

ASSIGNMENT REVIEW

Dissection and discussion of sample solutions. Exploration of main difficulties.

#SESSION 12 (LIVE IN-PERSON)

“VOICES FROM…” SEMINAR SERIES: INDUSTRY

Seminar given from a leading figure in the industry. Review of exemplar use cases for ML in the industry. Question time and debate.

#SESSION 13 (LIVE IN-PERSON)

MODEL EVALUATION

Extended analysis of validation metthods. Cross-validation, Leave-one-out… Cases of use. Performance metrics and consistency. Bias-variance tradeoffs.

Book Chapters : Machine Learning Engineering (Chapter 7: Model Evaluation)

#SESSION 14 (LIVE IN-PERSON)

PERFORMANCE METRICS

Review and comparison of performance metrics. Supervised vs unsupervised. Accuracy, F-alpha score, Sensistiviy and Specificy. Generalization to multi-class and multi-label classification.

#SESSION 15 (LIVE IN-PERSON)

MODEL DEPLOYMENT AND MAINTENANCE - AN INTRODUCTION/OVERVIEW

Model life-cycle. Data and concept drift

Book Chapters : Machine Learning Engineering (Chapter 8: Model Deployment and Chapter 9: Model Serving, Monitoring, and Maintenance) (See Bibliography)

#SESSION 16 (LIVE IN-PERSON)

NEURAL NETWORKS AND DEEP LEARNING

Introduction to perceptron and multi-layer perceptron. Inner workings and mathematical foundation of a neural network. Forward and Backward propagation. Automatic differentiation. Gradient descent and its adaptative variants. Main hyperparameters. Dropouts and other regularization techniques.

#SESSION 17 (LIVE IN-PERSON)

NEURAL NETWORKS. REPRESENTATION LEARNING

Representation learning techiques. Conceptual explanation through practical examples. Representational models. Typical well-known architectures and their uses.

#SESSION 18 (LIVE IN-PERSON)

PRACTICAL SESSION

Assignment III: Coding a Neural Netwrok

How to code a NN. Basic concepts: tensors and gradient tapes. Activation functions and Optimizers. General design principles.

#SESSION 19 (LIVE IN-PERSON)

REINFORCEMENT LEARNING. AN INTRODUCTORY OVERVIEW

Introduction to intelligent agents based on the state-action-reward paradigm. Temporal differences and Bellman optimality equation. From iterative value functions optimization to policy gradients algorithms. Analisys of a practical example case of use.

#SESSION 20 (LIVE IN-PERSON)

SEQUENTIAL MODELING. FROM TIME SERIES ANALYSIS TO DEEP NN MODELS

Specific issues in time/sequential data feature engineering. Basic techniques involved. Example- based analysis of the different approaches and principles for sequential modeling and their evolution in the Machine Learning field.

#SESSION 21 (LIVE IN-PERSON)

TRANSFER LEARNING AND CONTRASTIVE LEARNING. THEORETICAL BASIS AND PRINCIPLES

Shallow training, principles and concepts and introduction to basic fine-tuning techniques. Perspectives for transfer learning multi-task learning

#SESSION 22 (LIVE IN-PERSON)

EXTENDED PRACTICE. CASE REVIEW

Assignment: GROUP ASSIGNMENT

Presentation of the Group Assignment.

#SESSION 23 (LIVE IN-PERSON)

“VOICES FROM…” SEMINAR SERIES: DEVELOPMENT

Seminar given from a leading ML developer. Review of approaches, tools and best practices. Question time and debate.

#SESSION 24 (LIVE IN-PERSON)

GROUP PRACTICE

#SESSION 25 (LIVE IN-PERSON)

REAL WORLD BUSINESS APPLICATIONS. OTHER FORMS OF LEARNING. CHALLENGES AND RISKS. STEPS AHEAD

Analysis of current research in the field and expected breakthoughs. Summary of risks: from adversarial attacks to mesa-optimization disalignment.

Book Chapters : Machine Learning Engineering (Chapter 10: Conclusion) (See Bibliography)

#SESSION 26 (LIVE IN-PERSON)

GROUP PRACTICE. CASE REVIEW

Dissection and discussion of sample solutions. Exploration of main difficulties.

#SESSION 27 (LIVE IN-PERSON)

“VOICES FROM…” SEMINAR SERIES: RESEARCH

Seminar given from a leading figure in the academic research. Using ML in cutting edge scientific research. Question time and debate.

#SESSION 28 (LIVE IN-PERSON)

FINAL EXAM

Concept summary review examination

#SESSIONS 29 & 30 (LIVE IN-PERSON)

GROUP PROJECTS - PRESENTATIONS