École Nationale Supérieure de
Techniques Avancées

Abstract

Predicting the stock price of gold for the next month is a common dream for many shareholders. However, just like the meteorologist and his weather report, it's a safe bet that the forecasts will prove to be wrong beyond a few weeks. At the heart of our daily life, Chaos is a recent eld of mechanics which governs a large number of systems, sometimes at the heart of our concerns as is the case with climate. This report rst seeks to enlighten the reader on the denition of Chaos and its most notable manifestations. Subsequently, elements of understanding related to the construction of neural networks and deep learning are provided. A discussion then opens on the results obtained from neuronal prediction models. Last but not least, a demonstration is made on the importance and benets of including physical sense in the construction and training of neural networks.