Program

The school is organized through three different initiatives:

Minicourses:

  • Albert Diaz Guilera, Universitat de Barcelona (10 hours): Complex networks foundations
  • Tomaso Aste UCL, London (6 hours): Information filtering networks for socio-economic systems
  • Matteo Matteucci, Politecnico di Milano (10 hours): Introduction to neural networks: from theory to practice
  • Josef Teichmann, ETH Zurich (6 hours) Provable Machine Learning Techniques in Finance

Lectures:

  • Christoffer Kok, European Central Bank, Contagion modelling at the ECB: analytical frameworks and policy usage
  • Paolo Giudici, Università di Pavia, Explanable AI credit risk models for peer to peer lending.
  • Andrea Prampolini, Intesa Sanpaolo, Limit order book simulation with interactive agents
  • Giuseppe Bruno, Banca d’Italia, Anomaly Detection in RTGS Systems: Performance Comparisons Between Shallow and Deep Neural Networks
  • Michele Tumminello, Università di Palermo, Insurance fraud detection: a statistically validated network approach
  • Daniele Marazzina, Politecnico di Milano, A machine learning model for lapse prediction in the life insurance contracts
  • Marcello Restelli, Politecnico di Milano, Reinforcement Learning for Automated Trading

Workshops: two workshops with researchers (junior and senior) debating research ideas.

Programme

Monday 14 June Tuesday 15 June Wednesday 16 June Thursday 17 June Friday 18 June
9.00-12.30 Matteucci Matteucci Aste Diaz Aste
13.00-14.00 Marazzina Restelli Giudici Workshop Matteucci/Diaz Research workshop
14.30-18.00 Diaz Diaz Matteucci Teichmann Teichmann
18.30 Prampolini Bruno Kok Tumminello

Albert Diaz Guilera (10 hours)

Bibliography:

Complex networks foundations
1st Lecture: Introduction
2nd lecture: Local characterization
3rd Lecture: Global characterization
3rd lecture: Models (Erdos-Renyi, Watts-Strogatz, Barabasi-Albert, …)
4th Lecture: Communities and more complex structures
5th Lecture: Dynamics

Filippo Menczer, Santo Fortunato, Clayton Davis (2020) A First Course in Network Science, Cambridge University Press.


Matteo Matteucci (10 hours)

Introduction to neural networks: from theory to practice

– Intro to neural networks & Feed forward neural networks (2h)
– Learning via backpropagation (while dealing with overfitting) (2h)
– Intro to Keras and neural networks coding (2h)
– Recurrent Neural Networks, vanishing gradient e long short-term memories (2)
– Recurrent neural networks training in Keras (2h)
– Deep Learning introduction (2h)

Bibliography:
Christopher M. Bishop, Neural Networks for Pattern Recognition, Oxford University Press, 1996
Ian Goodfellow, Yoshua Bengio, Aaron Courville, Deep Learning, The MIT Press, 2016


Tomaso Aste (6 hours course)

Information filtering networks for socio-economic systems

We live in a complex world where information is hard to process and the consequences of our decisions are difficult to predict. Humans, have learned to navigate quite efficiently this complexity. Mathematics, statistics and machines are far less capable. I will show how information filtering networks can help to understand and model this complexity.

In this lectures we will learn:

  1. what information filtering networks are
  2. how can we can build them
  3. how can we use them for characterization of socio-economic systems
  4. how can we use them for probabilistic modeling
  5. how can we apply them to stress testing, investment strategies and portfolio construction.

Pseudo-codes  will be introduced and examples in matlab and/or python will be shared and run in class to learn form examples how to build information filtering networks such as PMFG, TMFG, MFCF [1,2,3] and how combine them with graphical modeling methodologies (LoGo) [4] and tackle problems of practical relevance [5,6].

Bibliography:
[1] Tumminello, Michele, Tomaso Aste, Tiziana Di Matteo, and Rosario N. Mantegna. “A tool for filtering information in complex systems.” Proceedings of the National Academy of Sciences 102, no. 30 (2005): 10421-10426.

[2] Massara, Guido Previde, Tiziana Di Matteo, and Tomaso Aste. “Network filtering for big data: Triangulated maximally filtered graph.” Journal of complex Networks 5, no. 2 (2016): 161-178.

[3] Massara, Guido Previde, and Tomaso Aste. “Learning clique forests.” arXiv preprint arXiv:1905.02266 (2019).

[4] Barfuss, Wolfram, Guido Previde Massara, Tiziana Di Matteo, and Tomaso Aste. “Parsimonious modeling with information filtering networks.” Physical Review E 94, no. 6 (2016): 062306.

[5] Procacci, Pier Francesco, and Tomaso Aste. “Forecasting market states.” Quantitative Finance 19, no. 9 (2019): 1491-1498.

[6] Aste, Tomaso. “Stress testing and systemic risk measures using multivariate conditional probability.” Available at SSRN 3575512 (2020).


Josef Teichmann, ETH Zurich (6 hours)

Provable Machine Learning Techniques in Finance

We shall introduce two instances of Machine Learning in finance, Deep Hedging and Deep Simulation, and all the underlying mathematical concepts like a universal approximation on sigma compact spaces, stochastic search algorithms, and randomized signature methods. Theory will be accompanied by coded examples.