École d’été 2022

4–7 septembre 2022

Sebastian Engelke

Prof. Sebastian Engelke, UNIGE

Mme Caroline Gillardin, coordinatrice CUSO

Co organisation avec l'EPFL et l'EPFZ

Prof. Bin Yu from University of California, Berkeley (USA)

Prof. Johannes Schmidt-Hieber from University of Twente, Enschede (The Netherlands)

Prof. Aaditya Ramdas, Université Carnegie Mellon, Pittsburgh (USA)

You will find on our web site the slides of our speakers on the following link (please note we will send you the password by e-mail separately):

https://drive.switch.ch/index.php/s/Ld09bQucg9wXF0a

3 cours de 4 à 5h30 chacuns seront donnés par des spécialistes internationaux sur des sujets de pointes en statistiques.

Prof. Bin Yu de University of California, Berkeley (États-Unis)

Title: Veridical data science and interpretable machine learning towards trustworthy AI

Abstract

Lecture 1:This lecture introduces the predictability computability stability (PCS) framework and documentation that unifies, streamlines and expands on ideas and best practices from statistics and machine learning for the entire data science life cycle.
Lecture 2:This lecture discusses a motivating application of PCS to develop iterative random forests (iRF) that adds appropriate stability to random forests (RF) for discovering predictable and interpretable high-order interactions. iRF is illustrated through interdisciplinary research in genomics and medicine.
Lecture 3:This lecture first introduces a definition of interpretable machine learning through predictive accuracy, descriptive accuracy and relevancy to a human audience and a particular domain problem. Then it discusses methods such as ACD and AWD to interpret deep neural networks towards trustworthiness, in general and in the context of scientific collaborations in cosmology and cell biology

Prof. Johannes Schmidt-Hieber from University of Twente, Enschede (The Netherlands)

Title: Statistical theory for deep neural networks

Abstract

Lecture 1:Survey on neural network structures and deep learning. There are various types of neural networks that differ in complexity and the data types that can be processed. This lecture provides an overview and surveys the algorithms used to fit deep networks to data. We discuss different ideas that underly the existing approaches for a mathematical theory of deep networks.
Lecture 2: Theory for shallow networks . We start with the universal approximation theorem and discuss several proof strategies that provide some insights into functions that can be easily approximated by shallow networks. Based on this, a survey on approximation rates for shallow networks is given. It is shown how this leads to statistical estimation rates. In the lecture, we also discuss methods that fit shallow networks to data.
Lecture 3: Statistical theory for deep networks. Why are deep networks better than shallow networks? We provide a survey of the existing ideas in the literature. In particular, we study localization of deep networks and specific functions that can be easily approximated by deep networks. We outline the theory underlying the recent bounds on the estimation risk of deep ReLU networks. In the lecture, we discuss specific properties of the ReLU activation function. Based on this, we show how risk bounds can be obtained for sparsely connected ReLU networks. At the end, we describe important future steps needed for the further development of the statistical theory of deep learning.

Practical:The participants should try whether they can run the attached source code prior to the tutorial (we are not allowed to send python code via email. I have therefore changed the file ending to .txt. The file ending should be changed back from .txt to .py). The code has been written by Thijs Bos (University of Leiden). We recommend to install Anaconda and use spyder. Additional packages have then to be installed via Anaconda. Most notably Keras/tensorflow does not work with all version of pythons and it can be a bit tricky to install it. During the tutorial, we will also ask the participants to work a bit on their own with this program along some questions that we prepare.

Prof. Aaditya Ramdas, Université Carnegie Mellon, Pittsburgh (USA)

Title: game-theoretic statistics and safe anytime-valid inference

Abstract

Lecture 1 :Estimating means of bounded random variables by bettingWe introduce the principle of "testing by betting" by discussing in detail a simple and classical problem from probability theory and statistics from the 1960s (Hoeffding): given observations from a bounded distribution, how can we estimate its mean? This is a nonparametric estimation problem, for which we present state-of-the-art confidence intervals and "confidence sequences", that were derived from a decidedly game-theoretic perspective. We will introduce "Ville's inequality", the central (and only) mathematical inequality that underlies game-theoretic statistics. Time permitting, we will discuss some history, and what exactly Ville established in his seminal 1939 PhD thesis.

Lecture 2: Sequential experimental design: the lady tasting tea and universal inference. One of the main ideas in game-theoretic statistics is that we bet against the null, and directly use the resulting "wealth as evidence against the null". The resulting evidence processes are called "e-processes" (or e-values) and have many advantages over traditional notions of evidence like p-values. Primarily, if the evidence does not suffice, one can extend the experiment for free and collect even more evidence while maintaining a strong notion of error control. We will elaborate on these ideas by revisiting Fisher's classical "lady tasting tea" experiment from 100 years ago from a new, modern lens, and discuss extensions to multiple testing. If time permits, we will give a general methodology for constructing e-processes: universal inference.

Lecture 3 : Betting for sampling without replacement: how to audit elections. We discuss a very different, but interesting, application of testing and estimation by betting: how to audit an election (after an election result has been announced). This boils down to answering questions about sampling without replacement, another very well studied and classical problem for which betting provides a powerful new answer. Time permitting, we will end with an overview of other topics in nonparametric statistics and probability theory that were not covered in these three lectures.

 This program may be changed

Saignelégier - Centre de Loisirs des Franches-Montagnes - Jura

Centre de loisirs des franches-montagnes (CLFM) (JURA-RESORT)

Adress : Chemin des Sports 10, 2350 Saignelégier, Canton du Jura

Tel +41 (0)32 951 24 74

L'Hôtel Cristal est situé dans le centre de loisirs (CLFM)
Chemin des Sports 10, 2350 Saignelégier (JU)

www.clfm.ch

Accès à Saignelégier en voiture: Google Map

EN AVION Aéroports internationaux de: - Genève (180 km) - Zürich (145 km) - Bâle (78 km)

Horaires Swiss Train :

De: Aéroport de Genève : Train direct jusqu'à Neuchâtel. Ensuite train de Neuchâtel à La Chaux-de-Fonds. Ensuite train de la Chaux-de-Fonds à Saignelégier. Durée des trajets: Genève - Neuchâtel (69 minutes), Neuchâtel - La Chaux-de-Fonds (28 minutes), La Chaux-de-Fonds - Saignelégier (34 minutes).

Visa pour la Suisse (demande de visa suisse en ligne)

Météo en suisse (meteoswiss.admin.ch)

Tarifs si l'activité a lieu en présentiel :

Doctorant CUSO chambre double: 200 CHF

Doctorant CUSO chambre simple: 350 CHF

Post-doctorant CUSO chambre double: 300 CHF

Post-doctorant CUSO chambre simple: 450 CHF

Professeur CUSO chambre double: 400 CHF

Professeur CUSO chambre simple: 550 CHF

Non CUSO universitaire chambre double: 850 CHF

Non CUSO universitaire chambre simple: 1000 CHF

Non CUSO privé chambre double: 1300 CHF

Non CUSO privé chambre simple: 1500 CHF

Lors de votre inscription, merci de bien vouloir indiquer dans la zone commentaire si vous désirez une chambre simple, ou double et le nom de la personne avec qui vous souhaiteriez partager votre chambre. Dans le cas où rien n'est indiqué, une chambre simple sera réservée.

Versement sur compte postal (seulement si l'activité à lieu en présentiel):

CUSO
CCP 12-1873-8
Neuchâtel
BIC : POFICHBEXXX
IBAN : CH0509000000120018738. Merci d'écrire votre nom suivi du no » 22220001 " lors du paiement. Merci d'écrire votre nom sur le libellé du paiement et nb"22220001 ».

IMPORTANT :Si vous vous inscrivez et que vous n'êtes pas en mesure d'y assister, veuillez nous contacter dès que possible. Nous appliquerons la politique d'annulation de l'hôtel en cas d'information tardive.

35