Neuropsychology - Data Science

• eachers

  Bastien Blain, Clément Gorin and Paul Grignon

  Overview

  The goal of this course is to show the diversity of data sources and methods to model and analyze data mainly to behavioral and cognitive phenomena, but not only.

  It aims at providing students with:

  • Main concepts in Data Science regarding data exploration, modeling, analysis, and interpretation.
  • Tools and use cases in to translate concepts learnt into concrete applications. Most of the sessions will thus rely on a combination of theory and applications (lab sessions and concrete examples).

  A strong emphasis will be placed on applications. Notebooks with relevant pieces of code will be provided whenever relevant.

  Table of contents

  PART 1: RETRIEVING AND COLLECTING DATA

  In a first part, the course will deal with the wide range of data types available (e.g. textual

  data, web-related and social media activity data). It will also allow studying how existing data

  and new data can be mobilized through an experimental approach. It will provide some

  concrete use cases and tools to allow students to mobilize these diverse sources of data after

  the class.

  Session 1 / Using existing data to study cognitive and behavioural phenomena / Paul Grignon

  1. A diversity of existing data sources

  1.1. Diversity of data sources and associated tools to retrieve them.

  1.2. Example: a sentiment analysis task using social network data.

  2. Using existing data: the case of natural and quasi-natural experiments

  
  

  PART 2: DATA SCIENCE FROM DATA EXPLORATION TO MODEL INTERPRETATION

  In a second part, the course will provide an overview of the main concepts and tools in Data

  Science, in order to lead a successful data exploration, choose the suited modeling approach,

  properly implement and evaluate models, and interpret the resulting outputs.

  
  

  Session 2 / From exploration to modeling and analysis: main concepts and tools / Paul Grignon

  1. Data exploration : guidelines and recommendations

  2. Data Science

  2.1. Definition and scope

  2.2. A mapping of approaches and models

  3. Zoom: introduction to Machine Learning

  3.1. Main concepts

  3.2. Steps to implement a Machine Learning model

  
  

  Session 3 / Collecting new data through experiments: from design to analysis / Bastien Blain
  

  Hypothesis testing is key to behavioural science data science. Yet, not all data can be used to test all kind of hypothesis. A crucial -yet often neglected- step is to test whether your design allows you to test for your hypotheses. In this part of the class, we will see how to use simulations to run parameter and model recovery analyses, as well as potential tools to deal with data when different dimensions are correlated. 

  1. Principles of design optimization, model fit through numeric simulations

  2. Application to decision under risk, intertemporal choices, and reinforcement learning

  
  

  PART 3: FOCUS ON SOME APPROACHES

  Based on the key concepts and tools learnt in part 2, part 3 will allow applying the different

  concepts and tools through focusing on different useful modeling approaches when dealing

  with data.

  Session 4 / Focus on neural networks and the use of AI to analyze data / Clément Gorin

  1. Feed-forward networks

  1.1. Neurons and layers

  1.2. Optimization (backpropagation)

  1.3. Learned representations

  Session 5 / Focus on neural networks and the use of AI to analyze data / Clément Gorin
  

  2. Applications

  2.1. Overview of applications

  2.2. Implementation (TensorFlow, TBD)

  
  

  Session 6 / Other approaches to better understand and predict behavioural phenomena / Paul Grignon

  1. Using information contained in texts : Natural Language Processing

  2. Taking into account individual heterogeneity

  1.1. Beyond the average effect

  1.2. Clustering and segmentation

  2. Analysing sequential data : sequences of choice and decision processes

• References.

  Here are a few references that could be studied before the class. An enriched list will be provided to students enrolled to the class gradually during sessions.

  Books :

  Buisson, F. (2021). Behavioral Data Analysis with R and Python. " O'Reilly Media, Inc.".

  Hastie, T., Tibshirani, R., Friedman, J. H., & Friedman, J. H. (2009). The elements of

  statistical learning: data mining, inference, and prediction (Vol. 2, pp. 1-758). New York:

  springer.

  Reports :

  Report by the Behavioural Insights Team « Using Data Science in Policy » : https://www.bi.team/wp-content/uploads/2017/12/BIT_DATA-SCIENCE_WEB-READY.pdf

  Grading.

  The grading will take into account:

  •  Attendance – 10%.
  • Case study (final exam) – 40%.
  • Data Science project (lead by group of students) relying on an applied behavioural question (homework exam) – 50%.

Informations sur l'espace de cours

Rattachements à l'offre de formation