AI/ML - 2nd Advanced training week

Dates

13/03/2023 - 16/03/2023

Location

Florence

Level

Advanced training

Delivery mode

Residential

Course materials

It is hard to name a sector that will not be dramatically affected by machine learning or even artificial intelligence. Many excellent courses teach you the mechanics behind these innovations, helping you to develop an engineering skill set. This course takes a different approach. It is aimed at people who want to deploy these tools in business or policy, whether through start-ups or within larger organisations. While this requires some knowledge of how these tools work, it is only a tiny part of the equation — just as knowing how an engine works is a small part of understanding how to drive. What is needed is an understanding of what these tools do well and what they do poorly. This course focuses on giving you a functional, rather than mechanical, understanding. By the end, you should be an expert at identifying ideal use-cases and be well-equipped to improve analysis and policy using machine learning.
This course thus aims to give an overview and a grasp of popular Machine Learning (ML) techniques. Within this, you will be provided with an understanding of the opportunities and limitations of said techniques and a chance to interact with the experts. In practical terms, you will work hands-on with ML/AI methods in Python and demystify the black boxes to continue learning by yourself. Furthermore, the course will also present and discuss current policy and industry practice at European and international levels.

We were given an excellent overview of general AI/ML use cases and projects in European supervisory context. The course structure allowed us to learn about many different topics, try them instantly with hands-on experience and, most importantly, meet and exchange ideas with various experts from the supervision authorities all over Europe.

Karel Scheib

CNB - Ceska Narodni Banka (Czech National Bank)

Measure and evaluated sustainability, accuracy, fairness and explainability of AI (SAFE)
Analyse and compare the difference between AI and sample regressions models
Develop database for AI and ML applications
Estimate decision tree and random forest models
Implement Natural Language Processing applications
Discuss ESAs and NCAs SupTech best practices and case of applications
Design qualitative SupTech applications (Micro prudential, Macroprudential, consumer protection, market conduct supervision or AML/CFT)
Design quantitative SupTech applications (Micro prudential, Macroprudential, consumer protection, market conduct supervision or AML/CFT)

Part one

Lecture 1: Course overview
Tutorial 1: How to read data and use sklearn?
Lecture 2: Introduction to Machine Learning (ML)
Tutorial 2: Regressions versus Classifiers
Lecture 3: Machine Learning – the basics
Tutorial 3: Data pre-processing and assessing model performance
Lecture 4: Fighting the curse of dimensionality
Tutorial 4: Cross-validation applied to LASSO variable selection
Lecture 5: Improving weak learning
Tutorial 5: Decision trees and random forests
Lecture 6: Unsupervised learning and clustering
Tutorial 6: Finding clusters and neighbours
Lecture 7: Natural Language Processing (NLP)
Tutorial 7: NLTK and sentiment analysis
Lecture 8: Explainability

From practice to policy: AI ACT supervisor implication

Part Two

Session 1: NCAs use case of Suptech applications
Session 2: Future trends, challenges, and collaborations in Suptech: ESAs prospective
Session 3: Technology application design and prototyping
Session 4: Industry dialogue: future market development

Background in statistics, computer science, economics, finance, or related fields
Basic knowledge of programming languages (Python, R, C++, etc.)

AI/ML – 2nd Advanced training week