Carlo Ciliberto

PhD in Robotics
Professor in Machine Learning
University College London, UK

Areas of Expertise:
Machine Learning
Statistical Learning
Computer Vision

Research:
Carlo Ciliberto research interests focus on foundational aspects of machine learning within the framework of statistical learning theory. He is particularly interested in the role of “structure” (being it in the form of prior-knowledge or structural constraints) in lowering the sample complexity of learning algorithms. He investigated these questions within the settings of structured prediction, multi-task and meta-learning, with applications to computer vision and humanoid robotics.

2021-Present     Associate Professor, Computer Science, University College London, UK.
2018-2021         Lecturer, Electrical and Electronic Engineering, Imperial College London, UK.
2017-2018         Research Associate, Computer Science, University College London.
2012-2017         Postdoctoral Fellow, Massachusetts Institute of Technology, USA.
2009-2012         PhD student in robotics, Istituto Italiano di Tecnologia, Italy.
2006-2008         Laurea Specialistica in Matematica, Roma Tre, Italy.
2003-2006         Laurea Triennale in Matematica, Roma Tre, Italy.

Sylllabus

Statistics and Optimization for Machine Learning
This course will introduce students to the key ideas in modern machine learning from a principled mathematical perspective. We will understand the role of the abstract concept of regularization in tackling overfitting and how the delicate process of model selection is critical to properly apply it in practice. We will study the main algorithms from optimization that enable data scientists to train a machine learning model. The module will conclude with an in-depth analysis of the main applications of these tools in practice.
 
Intended Learning Outcomes:

By the end of this module, students will be able to:

  • Understand the mathematical definition of the concept of “learning”.
  • Implement state-of-the-art machine learning algorithms from scratch.
  • Explain the key differences between a wide range of learning algorithms.
  • Choose the most suited algorithm for a given task.
  • Tackle the delicate process of model selection from a principled perspective.
  • Act to fix a learning algorithm when it does not work in practice.

Topics:

  • A taxonomy of Machine learning,
  • Overfitting and Regularization,
  • Local Methods,
  • Linear Models,
  • Kernel Methods,
  • Model Selection,
  • Optimization for Machine Learning,
  • Early Stopping,
  • Feature Selection,
  • Big data and Large-scale learning,
  • Multi-task and Meta-learning.
  • Structured Prediction.