Data Science Workshops

Barcelona GSE Data Science Workshops are sessions conducted by experts on a specific topic. The content and format are designed to be interesting for professionals and academics alike. Workshops last 4-6 hours and are held at Ciutadella Campus (UPF) in Room 24.S19. 

Special pricing is available for professors, PhD students, and Barcelona GSE Alumni. Current students in the Barcelona GSE Master's programs may attend the workshops for free. Registration is required.

Upcoming Barcelona GSE Data Science Workshops

Workshop on Transportation in Machine Learning
Anestis Papanikolaou (JACOBS Urban Mobility, London)
MAY 4 & 7 | 17:00-20:00

The main objective of this course is to provide a thorough grounding on the key principles of transportation planning and modelling. The theoretical underpinnings of transport science have evolved from many disciplines, most notably economics, statistics, psychology and operational research. Transportation planning and modeling is currently in a transformative phase, as the emergence of new data sources as well as the use of machine learning techniques open up a new spectrum of opportunities for transportation researchers and practitioners.

The course includes a complete introduction to transportation theory (passenger transport) in order to equip students with the fundamental concepts, assumptions, models and algorithms met in transport science and practice. Students will also have the opportunity to apply these methodologies in the context of real world problems met in transportation practice. The final objective is to familiarize students with the appropriate domain knowledge in order to be able to pursue a research and/or professional position in the evolving transportation sector by conveying the data scientist mindset

About the speaker

Anestis Papanikolaou is Principal Transport Modeller for JACOBS Urban Mobility in London.

Workshop on Reinforcement Learning
Gergely Neu (DTIC, Universitat Pompeu Fabra)
May 8 & 9 | 17:00-20:00

Prerequisites: Statistical modelling and inference, Machine Learning, Stochastic models and optimization

Reinforcement learning (RL) is the model-based theory of sequential decision-making under uncertainty. The field of reinforcement learning research looks back to a long history with several periods of high and moderate success; recent years saw an unprecedented increase in attention to the field both in the mainstream media and within the machine learning/AI research community. Indeed, RL techniques underly many of the breakthrough successes of AI research that made headlines in recent years (e.g., the ATARI and Go players developed at Google DeepMind). Within the machine learning community, several successful and influential researchers now cite the task of general reinforcement learning as the ultimate challenge for all of machine learning research. Very recently, RL has been highlighted as one of the "Top 10 Breakthrough Technologies of 2017" by the MIT Technology Review.

This tutorial gives an overview of the fundamental techniques of reinforcement learning starting from the classic temporal-difference methods through approximate dynamic programming all the way to recent developments in deep reinforcement learning. The goal is to provide a strong understanding of the most common methods and provide a basic algorithmic toolbox for building RL systems. Besides familiarizing students with these tools, the course puts a strong emphasis on highlighting the crucial challenges that set RL problems apart from other machine learning problems. Students taking the course are expected to gain the capability to identify and tackle such challenges in various application domains.

About the speaker

Gergely Neu is a postdoctoral researcher at Universitat Pompeu Fabra. He has previously worked with the SequeL team of INRIA Lille, France and the RLAI group at the University of Alberta, Edmonton, Canada. He obtained his PhD degree in 2013 from the Technical University of Budapest, where his advisors were Andras Gyorgy, Csaba Szepesvari and Laszlo Gyorfi. His main research interests are in machine learning theory, including reinforcement learning and online learning with limited feedback and/or very large action sets.

Workshop on Distributed Machine Learning
Ioannis Arapakis (Telefónica Research)
JUNE 4 & 6 | 17:00-20:00

I​​n this workshop we will examine the basic concepts (RDDs, transformations, runtime architecture) behind distributed Machine Learning (ML) using Spark, a fast and general-purpose cluster computing platform. We will motivate the use of distributed ML and introduce MLlib, a library of machine learning functions. MLlib has been designed to run in parallel on clusters, contains a variety of learning algorithms and is accessible from all of Spark’s programming languages (e.g., Java, Python).

During the lab session we will try to implement some basic supervised/unsupervised algorithms introduced in the lecture. We will aim to be using python and some additional useful libraries, such as NumPy and Pandas.

About the speaker

Ioannis Arapakis is a Research Scientist at Telefónica Research in Barcelona.

Previous workshops

Workshop on Causality
Jonas Peters (University of Copenhagen)

In the field of causality we want to understand how a system reacts under interventions (e.g. in gene knock-out experiments). These questions go beyond statistical dependences and can therefore not be answered by standard regression or classification techniques. In this tutorial you will learn about the interesting problem of causal inference and recent developments in the field. No prior knowledge about causality is required.

Part 1: We introduce structural causal models and formalize interventional distributions. We define causal effects and show how to compute them if the causal structure is known.

Part 2: We present three ideas that can be used to infer causal structure from data: (1) finding (conditional) independences in the data, (2) restricting structural equation models and (3) exploiting the fact that causal models remain invariant in different environments.

Part 3: If time allows, we show how causal concepts could be used in more classical machine learning problems.

About the speaker

Jonas Peters is Associate Professor in Statistics at the University of Copenhagen.

Workshop on Deep Learning
Alexandros Karatzoglou (Telefónica Research)

In the field of causality we want to understand how a system reacts under interventions (e.g. in gene knock-out experiments). These questions go beyond statistical dependences and can therefore not be answered by standard regression or classification techniques. In this tutorial you will learn about the interesting problem of causal inference and recent developments in the field. No prior knowledge about causality is required.

The course will offer an introduction to deep learning along with an extensive practical hands-on session in python. In the course we cover deep feedforward models, convolutional networks used mainly in image processing, recurrent neural networks used commonly in text processing, autoencoders, word2vec and more. The course aims to introduce also elements of optimization for deep learning. The hands-part of the course will focus on the use of deep learning technics on Image and text data in particular natural language processing.

About the speaker

Alexandros Karatzoglou is the Scientific Director at Telefonica Research.