Number of hours
- Lectures 9.0
- Projects -
- Tutorials -
- Internship -
- Laboratory works 15.0
- Written tests -
ECTS
ECTS 3.0
Goal(s)
The goal of this class is to introduce notions about artificial neural networks, and study how they can be used by implementing a neural network library with several features, including gradient descent optimization using the Momentum strategy or training regularization thanks to L2 penalization. This library is used to solve a supervised learning problem on financial data.
Responsible(s)
Patrick REIGNIER
Content(s)
Introduction to artifical neural networks: historical overview and first identified limitations
Perceptron and multilayer perceptron, universality
Training of neural networks, derivation of backpropagation equations
Optimizing gradient descent: Momentum, Nesterov, Adam...
Regularization techniques: L1? L2, Dropout...
Input standardization, Batch Normalization
Implementation of a neural network library in Java
None
Test
Evaluation : Projet (évaluation en continu et sur le rendu)
Resit : Examen oral (exposé, soutenance, etc..) (30mns)
N1: Project assessment
N2: Project assessment
Calendar
The course exists in the following branches:
- Curriculum - Financial Engineering - Semester 9
Additional Information
Course ID : WMMF9M41
Course language(s): 
You can find this course among all other courses.
Bibliography
Ian Goodfellow and Yoshua Bengio and Aaron Courville: Deep Learning
Charu C. Aggarwal: Neural Networks and Deep Learning