MADAS: Pizza, coffee breaks, agent-based modeling and statistical learning

My first term in MADAS at the Collegio Carlo Alberto is almost over. It is time to reflect on the past three months. A lot of you will end up on this page through a search engine, looking for pieces of information. Since information on the program is scarce, the goal of this blog post is to provide you with some details to help you in your application decision.

The “Master in Data Science for Complex Economic Systems” (MADAS) at the Collegio Carlo Alberto is a one-year postgraduate training in the city of Turin. It is a somewhat eclectic postgraduate because although it has ‘data science’ in its name, it has a second focus on ‘complex economic systems’.

In the year 2017-18 it had the following courses which I would classify in two categories; statistical learning and complexity economics. There were also introductory courses and contextual and applied courses which I do not elaborate upon. You can find the full course list here.

Statistical learning

I introduce you to the lecturers, provide you with some info on the evaluation method, some main topics, a personal note and if it’s mathematics-heavy – because that was an important concern for some people, including me.

Data Science

Taught by Paola Cerchiello who is professor at the University of Pavia. The contents of this course align with the marvelous ‘ISLR’ book (PDF). This course covers both parametric and non-parametric models and some text mining algorithms. The mathematical side is minimal, although often crucial in understanding the algorithms. Coding examples are done in R. Evaluation is a final project on a data set of your choice. This course was my personal favourite and perfectly correlated with my main interest field.

Some topics: linear, multiple, logistic, ridge, lasso regression; regularization, cross-validation, bootstrap, trees, latent dirichlet analysis

Machine Learning

This course is taught by three people:

  • Ruggero G. Pensa: Assistant professor in computer sciences at the University of Torino. Teaches basic clustering techniques; KNN, hierarchical clustering, …
  • Roberto Esposito: Professor in computer science at the University of Torino. Teaches Neural Networks. He follows the splendid guide that has been written by Michael Nielsen. Mathematics is minimal, however, if you don’t know calculus and linear algebra, you will have a really hard time understanding how neural networks work.
  • Mattia Cerrato: Bass player and doctoral student at the University of Turino. Is the guiding light during the exercise courses. Courses are on clustering and on neural networks. For neural networks, Google’s Tensorflow is used.

The evaluation is an exam that features a theoretical part; two questions in one hour and a coding part; two questions in two hours. The exam is rather hard. Although the questions an sich aren’t of a very high level, finishing the exam within the time frame is next to impossible.


This course is taught by Matteo Manera and Consuelo Nava. She also taught the introductory course on statistics. This course is in the second term and I have yet to follow it.

This is what the Collegio looks like inside.
This is what the Collegio looks like inside.

Complexity Economics

Agent Based Modeling

This course is taught by three professors:

  • Magda Fontana: Professor at the Department of Economics & Statistics at the University of Torino. Mainly teaches how to work in NetLogo. Unfortunately she had to cancel teaching activities because of health reasons. I hope she returns next year because she’s not only a kind and warm person but also a great teacher.
  • Pietro Terna: Retired professor in economics. Apparently a phenomenon in the city of Turin as he is like a living Wikipedia on modeling and computer science. His classes teach you some basic skills you might need sometimes if you want to proceed with agent-based modeling.
  • Matteo Morini: A polyglot and polymath, just as Pietro Terna, who teaches at Lyon and collects Phd’s as a hobby. He taught us the thought process behind an agent-based model.

This course was somewhat chaotic as the professors repeated each other’s material. Evaluation form is a paper and a small project. There should have been a small exam for Fontana’s course, but this will be transferred to the second term. Personally, I am still not convinced about practical usage of agent based models and papers that use the technique can’t really convince me. I wrote my term paper about this.

Lastly, there’s barely any mathematics in this course, if that’s any of your concern.

Network Science

This course is taught by Marton Karsai, a French-Hungarian computer scientist and – from what I heard – an authority in the field of network science. If you don’t know what network science is, it’s a scientific field that tries to map complexity through edges and nodes. It was very novel to me and a surprisingly interesting course. From a mathematical point of view, this course is quite intensive. Many concepts and phenomena are substantiated with rigorous mathematical proof. Work on your calculus if you want to understand everything. The exam was quite silly because you were allowed to bring notes.

There is also a course by Pietro Terna on the usage of agent-based modeling in the field of network science which requires a paper at the end of the semester.

Economics of Complex Systems

This course is taught by Marco Guerzoni. I still have to take it in the second term.

From the last week of August until September there are introductory courses on statistics and programming (in both R and Python).

Introductory courses


This introductory course is taught by Consuelo Nava. It’s basically impossible to go from set theory to maximum likelihood estimation in three weeks. Yet that’s what this course succeeds in. It is a really, really intensive introduction that covers the main topics in statistics if you follow a bachelor in economics, or any other science. The evaluation is a written exam with basic exercises that are not algebra-heavy, yet very clever to find out if you understand the syllabus.

Some topics: Set theory, probability, random variables, distributions, points estimation, confidence intervals, hypothesis testing, MLE, …

The guideline for this course is the introductory book by Mood & Graybill.


This course is taught by Paolo Racca. Equal time is spent on basic programming techniques in both Python and R. When I say ‘basic’, I mean really basic. It goes from variables to conditional statements to loops and classes in a couple of weeks. The evaluation is a group work that covers several programming exercises.


Do you recommend it?

Many of my fellow students are really happy with the lectures and the material. Although going from zero programming experience to programming a neural network and programming a clustering job with cross-validation in three months is pretty intensive, most of them seemed to have picked up the skills quite well.

  • If you have experience in statistics and no experience in programming, I think MADAS is something for you.
  • If you have experience in programming but not in statistics, this is also something for you.
  • If you have no experience in programming and no experience in statistics; prepare for a tough year.
  • If you have experience in programming and in statistics, I’m not really convinced that you should apply.

What can I study to prepare for the course?

  1. Calculus: A good book is “Introduction to mathematical economics” by Edward T. Dowling. Do not study this as you would for a calculus exam. It suffices to have a good understanding of the concepts.
  2. Linear Algebra: This book is really good.
  3. Programming in R and Python: There are plenty of online courses. DataCamp, a Belgian startup, will help you get the hang of it

Can I combine MADAS with a job?

I have been combining it with a part-time job in the digital sector where I can organize my hours on an individual basis and this has been going pretty well. So yes, it’s possible, but be prepared to spend a lot of times in front of a computer. There’s no room for recurring patterns in your work schedule as every week is different at MADAS. One week you have 24 hours of lectures and another only 10.

Is it a real degree?

Although the website (and the name) claims it is a master; it actually it is not. It does not count towards ECTS credits. It’s a postgraduate training program (in Italy – a “Diploma di Perfezionamento”). Be aware that this can affect you in more ways than you expect. For example, after applying I discovered it’s not a master’s degree, it was not recognized by the Belgian (Flemish) government, I could not get funding and could not request a legal “career break” (some labour market policy in Belgium). Nevertheless, the Collegio Carlo Alberto is a respected institution. There’s two kinds of people in Italy: those who have never heard of it and those who wrongfully assume I must be supersmart because I enrolled at the Collegio.

Anything else?

Yeah, administratively, it’s kind of a mess. The kind ladies at the administration don’t speak English very fluently. It’s not very easy to get something done. e.g. I have been waiting for 8 months for an invoice of the registration fees. From the day that I applied until the minute I walked into a classroom I didn’t know if it was real or not. I can assure you, it’s not a scam. And with enough patience, you will get the papers you need. I guess it’s just an Italy thing :-).

Lastly, the Collegio Carlo Alberto moved from Moncalieri to the center of Turin in september 2017. Most pictures you will find are from the old building.