• Aucun résultat trouvé

“Training Data Scientists for Aeronautics and Space”Data and Decision Sciences @ Supaero

N/A
N/A
Protected

Academic year: 2021

Partager "“Training Data Scientists for Aeronautics and Space”Data and Decision Sciences @ Supaero"

Copied!
2
0
0

Texte intégral

(1)

“Training Data Scientists for Aeronautics and Space”

Data and Decision Sciences @ Supaero

Professional experience

Gap year

1st & 2nd year

Core eng. courses Elective courses : - Markov chains - Applied optimization - Intro to Big Data ...

3rd year

Data and Decision Sciences program Internship

Supaero curriculum

2015-16: 19 students

2016-17: 30 students

2017-18: 53 students

Foundations in Decision Making

Decision Theory – Statistics – Optimization

Data Mining & Machine Learning

Advanced Statistics – Supervized, Unsupervized, Reinforcement Learning

Tools of Big Data

Databases – Programming – Spark

Digital Economy and Data Uses

Business models – Privacy – Data storytelling

Syllabus

Hackathon

3 days challenge – real-life data Industrial partners

2016 edition on image analysis with IRT – Saint Exupéry

Link to 2016 edition press release

Program head

Emmanuel Rachelson

e.rachelson@isae-supaero.fr

50 trainers, experts

from academia or

major companies

Internships

Dassault, Airbus, Sopra Steria, Air France, Bloomberg, Thales, start-ups,

(2)

Detailed syllabus

Foundations in Decision Making 70 hours of classes on:

● Decision Theory (multi-criteria, multi-sources, collective decision making, uncertainty modeling) ● Statistics (descriptive, exploratory, inference)

● Combinatorial Optimization (Complexity analysis, Mixed Linear Programming, Constraint

Programming, Graph-based Optimization)

● Stochastic Optimization methods

Data Mining & Machine Learning 70 hours of classes on:

● Advanced statistics (statistical modeling, multivariate statistics)

● Unsupervized and Supervized Learning Algorithms (k-means clustering, Hierarchical clustering,

Neural networks, Deep Learning, Naives Bayes classification, Gaussian Processes, Support Vector Machines, kernel methods, Boosting, Bagging, Random Forests, Statistical Learning Theory)

● Reinforcement Learning and Markov Decision Processes (model based and model-free, online,

offline, Learning Classifiers Systems, Monte-Carlo tree search) Tools of Big Data

40 hours of classes on:

● Databases (theory and practice) ● Functional Programming

● Practice of Python (C and Java are taught in the core courses in 1st and 2nd year) ● Spark environment

Digital Economy and Data Uses 30 hours of classes on:

● Business models in the digital economy ● Data security and privacy issues

● Dataviz and data storytelling

Hackathon

Counts as a 30 hour class.

Autonomy and agility on a real-world challenge in Data Science with industrial validation. Seminars

All year long, by academics and professional experts. Internship

6 months, with academic validation.

In parallel

Supaero 3rd year: choice of an independent applicative domain of study (aircraft, space systems,

embedded systems, mathematical modeling, energy, transport, environment, etc.) Master of Science in Operations Research or Applied Mathematics

Références

Documents relatifs

In order to be able to use a dot product as a similarity measure, we therefore first need to represent the patterns as vectors in some dot product space H , called the feature

Here we use a Support vector machines to build a machine learning model for the ECG dataset, using a portion of the data (80%) for training and the rest for testing the model

In this paper, we focus on SVM-based vehicle detection technique, where we can simplify vehicle detection problem to a classification problem by examining

This paper describes a novel method for improving classifi- cation of support vector machines (SVM) with recursive feature selection (SVM-RFE) when applied to cancer classification

To tackle this problem, we utilize Dempster-Shafer evidence theory to measure the uncertainty of the prediction output by deep neural networks and thereby propose an uncertain

We apply the random neural networks (RNN) [15], [16] developed for deep learning recently [17]–[19] to detecting network attacks using network metrics extract- ed from the

• pour chaque observation, on peut faire des prédictions avec les modèles provenant d’échantillon ne la contenant pas. On obtient l’erreur OOB associée au processus

The expectations of those update vectors guarantee the convergence of the learning process and, thanks in part to Theorem 1 stated just below, it is guaranteed with probability 1 −