Alessandro Temperoni: Second-order methods for deep learning

Alessandro Temperoni: Second-order methods for deep learning

Machine Learning Seminar presentation

Topic: Second-order methods for deep learning

Speaker: Alessandro Temperoni, Faculty of Science, Technology and Medicine; University of Luxembourg

Time: Wednesday, 2021.07.14, 10:00 CET

How to join: Please contact Jakub Lengiewicz

Abstract:

Second-order methods techniques are widely used in the field of optimization and have been recently applied to neural networks with impressive results. In this seminar, I will present two applications of second-order methods for deep learning.

Additional material:

Juan Pineda-Jaramillo: Exploring significant predictors of freight rail intermodal operation delays using causal Machine Learning

Juan Pineda-Jaramillo: Exploring significant predictors of freight rail intermodal operation delays using causal Machine Learning

Machine Learning Seminar presentation

Topic: Exploring significant predictors of freight rail intermodal operation delays using causal Machine Learning

Speaker: Juan Pineda-Jaramillo, Faculty of Science, Technology and Medicine; University of Luxembourg

Time: Wednesday, 2021.07.07, 10:00 CET

How to join: Please contact Jakub Lengiewicz

Abstract:

Delays in freight rail intermodal operations generate negative impacts on the railway industry, so identifying the causes associated to these delays is vital to mitigate operational risks. In this seminar, we will present a set of Machine Learning models that were trained to predict the delays in freight rail intermodal operations, and then the most suitable model was used to explore the significant predictors which cause those delays, using data from the National Railway Company of Luxembourg that connects Bettembourg with several EU countries.

Additional material:

GRACM 2021: Fracture across space and time: a journey through multiscale methods, model order reduction and machine learning

GRACM 2021: Fracture across space and time: a journey through multiscale methods, model order reduction and machine learning

Professor Bordas will give a talk the 07/07/21 from 2:45 pm to 3:15 pm at the 10th GRACM.

To register : click here

For the full program: click here

Abstract

Fracture across space and time: a journey through multiscale methods, model order reduction and machine learning

We review and connect in this paper two approaches to solve multi-scale fracture problems. Machine learning and model order reduction on the one hand and multi-scale methods on the other hand. 

We explain as didactically as possible how material complexity has led to the need for acceleration methods and discuss advances in model selection and error estimation for such problems. 

We show how model reduction/machine learning and standard multiscale methods both fails when dealing with localization problems occurring in fracture mechanics. 

We conclude by discussing the possibility of digital twins and make a parallel with medical simulation. 

 

Literature


A computational library for multiscale modeling of material failure
H Talebi, M Silani, SPA Bordas, P Kerfriden, T Rabczuk
Computational Mechanics 53 (5), 1047-1071

Bridging proper orthogonal decomposition methods and augmented Newton–Krylov algorithms: an adaptive model order reduction for highly nonlinear mechanical problems
P Kerfriden, P Gosselet, S Adhikari, SPA Bordas
Computer Methods in Applied Mechanics and Engineering 200 (5-8), 850-866

A partitioned model order reduction approach to rationalise computational expenses in nonlinear fracture mechanics
P Kerfriden, O Goury, T Rabczuk, SPA Bordas
Computer Methods in Applied Mechanics and Engineering 256, 169-188

Local/global model order reduction strategy for the simulation of quasi-brittle fracture
P Kerfriden, JC Passieux, SPA Bordas
International Journal for Numerical Methods in Engineering 89 (2), 154-179

Molecular Dynamics/XFEM Coupling by a Three Dimensional Extended Bridging Domain with Applications to Dynamic Brittle Fracture
H Talebi, M Silani, S Bordas, P Kerfriden, T Rabczuk
International Journal for Multiscale Computational Engineering 11 (6), 527-541


Automatised selection of load paths to construct reduced-order models in computational damage micromechanics: from dissipation-driven random selection to Bayesian optimization
O Goury, D Amsallem, SPA Bordas, WK Liu, P Kerfriden
Computational Mechanics 58 (2), 213–234

Quasicontinuum-based multiscale approaches for plate-like beam lattices experiencing in-plane and out-of-plane deformation
LAA Beex, P Kerfriden, T Rabczuk, SPA Bordas
Computer Methods in Applied Mechanics and Engineering 279, 348–378


Statistical extraction of process zones and representative subspaces in fracture of random composite
P Kerfriden, KM Schmidt, T Rabczuk, SPA Bordas
International Journal for Multiscale Computational Engineering 11 (3), 253-287

Guaranteed error bounds in homogenisation: an optimum stochastic approach to preserve the numerical separation of scales
D Alves Paladim, JP Moitinho de Almeida, SPA Bordas, P Kerfriden
International Journal for Numerical Methods in Engineering 110 (2), 103–132

What makes data science different? A discussion involving statistics2. 0 and computational sciences
C Ley, SPA Bordas
International Journal of Data Science and Analytics 6 (3), 167-175

Mathematical modelling and artificial intelligence in Luxembourg: Twenty PhD students to be trained in data-driven modelling
S Bordas, S Natarajan, A Zilian
ERCIM News 115, 39-40

Maciej Skorski: Approximating Hessians for Neural Networks

Maciej Skorski: Approximating Hessians for Neural Networks

Machine Learning Seminar presentation

Topic: Approximating Hessians for Neural Networks

Speaker: Maciej Skorski, Faculty of Science, Technology and Medicine; University of Luxembourg

Time: Wednesday, 2021.06.30, 10:00 CET

How to join: Please contact Jakub Lengiewicz

Abstract:

Second-order optimization techniques have been proven successful, but cannot be fully leveraged for neural networks due to infeasibility of hessian calculations. I will discuss an approximated way of computing the hessian for neural networks, which emerges from a careful analysis of the hessian chain rule and activation functions.

Additional material:

Dr. Katerina E. Aifantis: Mechanical Behavior of Cells and Biomaterials 06/30/21 at 11:00 am

Dr. Katerina E. Aifantis: Mechanical Behavior of Cells and Biomaterials  06/30/21 at 11:00 am

Legato Team (https://legato-team.eu/) has organised a seminar on the topic Mechanical Behavior of Cells and Biomaterials. I am writing this email to invite you all to attend this insightful seminar that will be held on June 30, 2021 at Room No: MNO-E01-0146030 (Maison du Nombre) at 11:00 am (Limited places available due to Covid restrictions). You can also attend the seminar through Webex as well as Youtube live (Links are given below).

The seminar will surely be helpful for all students as well as researchers, thus do participate and gain maximum benefit out of it. 

———————————————————————————————————————————————————————————————————————————

Introduction to Speaker:

 

Dr. Katerina E. Aifantis

Associate Professor,

Mechanical and Aerospace Engineering,

University of Florida

 

Dr. Katerina E. Aifantis received her PhD from the University of Groningen in 2005 at the age of 21, becoming the youngest PhD in the Netherlands. After a short post-doctoral period at Harvard/US and at Ecole des Mines of Paris/France, She became the youngest recipient of the European Research Council Starting/ERC Grant at the age of 24, which she carried out at Aristotle University of Thessaloniki and the University of Erlangen-Nuremberg, between 2008-2013. In 2013, she joined the University of Arizona as an Associate Professor, and since 2017 she has been an Associate Professor and Faculty Fellow at the Mechanical and Aerospace Engineering Department of the University of Florida, where she set up the Laboratory of Nanomaterials for Energy and Biological Applications. Her research primarily focuses on using solid mechanics for understanding materials behavior at the nanoscale, such as dislocation-grain boundary and dislocation-graphene interactions. In addition to basic science questions, she uses her theoretical and experimental insight to predict the most promising materials systems that can be used in various applications, ranging from next generation electrodes for Li-ion batteries, to bone regeneration scaffolds and bioacompatible electrodes for deep brain stimulation.

* When: Wednesday, June 30, 2021 
Time: 11:00 a.m. (Paris time)

* Where:

Room No MNO-E01-0146030Maison du Nombre, University of Luxembourg (Belval Campus)  (Limited places available due to Covid restrictions).  

Webex Meet room: https://unilu.webex.com/unilu/j.php?MTID=m1a56ac0cc40f6070f0496603684679c7

Youtube link: https://youtu.be/M8WQFBSJkbs

 

* Title: Mechanical Behavior of Cells and Biomaterials

 

* Abstract:

       The present talk will focus on understanding the mechanical behavior of cells and biomaterials. Tissue regeneration is an area which depends on the ability to develop materials that poses the appropriate combination of microstructure and stiffness that promote cell adhesion and differentiation. Heart patches are a most challenging case, as they need to be seeded with cardiomyocytes and then placed in the infracted area of the heart. New polymer blends of Poly(Glycerol Sebacate) (PGS) prepolymer and Polybutaline Succinate (PBS-DLA) are introduced as possible patches, as they increase the viability of myocytes, depending on their elastic modulus and porosity.

      In the second part, similar fibrous materials will be used to study cancer cell evasion. Collagen co-polymers will be used of different concentrations in order to change the porosity while glutaraldehyde will be employed to change the stiffness. Seeding the different samples with evasive cancer cells, allows to capture the interplay between tissue pore size versus stiffness in promoting cell migration.

       In the third part, similar experimental tools used in determining the mechanical properties of biomaterials, will be used to examine the deformation of cells. Particularly, atomic force microscopy will be employed to capture the effects that the common blood thinning medication, pentoxifylline, has on the elastic modulus of red blood cells. This was the first in vivo experiment of its kind, as the blood samples were taken from human subjects taking this medication. In continuing to a different setup, optical tweezers will be used to show that cells undergo plastic deformation in addition to viscoelastic deformation.

Workshop on “Image and Physics” on June 18, 2021, Center of Mathematical Morphology and Center of Materials

Workshop on “Image and Physics” on June 18, 2021, Center of Mathematical Morphology and Center of Materials

We are pleased to announce a workshop centered on “image and physics” on June 18, 2021, organized by the Center of Mathematical Morphology and Center of Materials, devoted to image analysis, deep learning, physics, and material science. The main objective of this event is to promote exchanges in an informal setting between researchers of MINES ParisTech working on image analysis and deep learning, on the one hand, and on physics and material science in general. You are welcome to send your contributions (title, authors, and abstract) to the organizers before June 4th.

We hope to see you in June,

François Willot et Henry Proudhon

18/06/2021 at 9:30.link : https://mines-paristech.zoom.us/j/92437426470?pwd=YWlIblpndkRRTlQyQXpZK1pJaXY4QT09
<https://mines-paristech.zoom.us/j/92437426470?pwd=YWlIblpndkRRTlQyQXpZK1pJaXY4QT09>

(ID de la réunion / Meeting ID : 92437426470 Mot de passe / Password :
280307)

Piergiorgio Vitello: Enhancing Public Transport Demand Information Using Crowdsourced Data

Piergiorgio Vitello: Enhancing Public Transport Demand Information Using Crowdsourced Data

Machine Learning Seminar presentation

Topic: Enhancing Public Transport Demand Information Using Crowdsourced Data

Speaker: Piergiorgio Vitello, Faculty of Science, Technology and Medicine; University of Luxembourg

Time: Wednesday, 2021.06.16, 10:00 CET

How to join: Please contact Jakub Lengiewicz

Abstract:

In the era of Advanced Public Transport Systems, new technologies have been introduced and deployed providing multiple sources of data that can be utilized for demand estimation and analysis. In this study, we investigate the potential of a crowdsourced dataset such as Google Popular Times to quantify the correlation of the demand of a transit line in conjunction with the level of attractivity of its station and the surrounding local businesses to understand to which extent the characteristics of the activities in the area can be associated with the passenger transference at the stops.

Additional material:

MMLDT-CSET 2021 NSF Fellowship Application and registration

MMLDT-CSET 2021 NSF Fellowship Application and registration

Mechanistic Machine Learning and Digital Twins for Computational Science, Engineering & Technology (MMLDT-CSET 2021)

An IACM Conference September 26-29, 2021 – Hyatt Regency Mission Bay, San Diego, USA

(website: https://mmldt.eng.ucsd.edu)  

Co-sponsors: IACM, ASCE-EMI, IUTAM, USNC/TAM, USACM

 

1. NSF Fellowship Application (https://mmldt.eng.ucsd.edu/fellowship)

The NSF Fellowships are available to promote awareness of mechanistic machine learning and digital twins and knowledge dissemination to all levels in education, including high school, undergraduate, and graduate levels. A portion of the fellowship will be reserved for underrepresented groups to promote equity, diversity, and inclusiveness.

This NSF Fellowship will cover the conference and short course remote registration fees. Only US Citizens and Permanent Residents are eligible for NSF Fellowships.

1). NSF Fellowship for 𝐠𝐫𝐚𝐝𝐮𝐚𝐭𝐞 𝐬𝐭𝐮𝐝𝐞𝐧𝐭𝐬 𝐚𝐧𝐝 𝐩𝐨𝐬𝐭𝐝𝐨𝐜 𝐫𝐞𝐬𝐞𝐚𝐫𝐜𝐡𝐞𝐫𝐬 to attend the MMLDT-CSET Conference and Short Course 2.

2). NSF Fellowship for 𝐡𝐢𝐠𝐡-𝐬𝐜𝐡𝐨𝐨𝐥 𝐭𝐞𝐚𝐜𝐡𝐞𝐫𝐬 𝐚𝐧𝐝 𝐚𝐝𝐦𝐢𝐧𝐢𝐬𝐭𝐫𝐚𝐭𝐨𝐫𝐬 to attend the MMLDT-CSET Conference and Short Course 1.

3). NSF Fellowship for 𝐮𝐧𝐝𝐞𝐫𝐠𝐫𝐚𝐝𝐮𝐚𝐭𝐞 𝐬𝐭𝐮𝐝𝐞𝐧𝐭𝐬 to attend the MMLDT-CSET Conference and Short Course 1.

 

2. Early Registration (https://mmldt.eng.ucsd.edu/register)

The Conference and Short Courses Early Registration is open until June 30, 2021. The conference will be organized with a hybrid format consisting virtual and on-site sessions in 44 Mini-symposia under 8 Tracks.

On-Site Attendance Includes:

Registration fee covers Ice Breaker Reception on Sunday, Pre-Banquet Reception and Banquet on Tuesday, Coffee Break Refreshments, Panel Discussions, on-site plenary lectures and technical presentations, and remote technical presentations.

Virtual Attendance Includes:

Registration fee covers live-streamed on-site plenary lectures and remote technical presentations.

 

Panel Discussions: 

Mechanistic Data Science for High School and Undergraduates STEM Education and Applications

 

Short Courses:

1). Mechanistic Data Science (MDS) for STEM Education and Applications

2). Mechanistic Machine Learning for Engineering and Applied Science

 

 Organizing Universities 

UC San Diego, Northwestern University, Stanford University, Brown University, Columbia University, and Arts et Métiers Institute of Technology of France

 

Feel free to contact the conference organizers with any questions or concerns: ceer.ucsd@gmail.com 

Juntong Chen: Robust estimation of a regression function in exponential families

Juntong Chen: Robust estimation of a regression function in exponential families

Machine Learning Seminar presentation

Topic: Robust estimation of a regression function in exponential families

Speaker: Juntong Chen, Faculty of Science, Technology and Medicine; University of Luxembourg

Time: Wednesday, 2021.06.02, 10:00 CET

How to join: Please contact Jakub Lengiewicz

Abstract:

In this talk, we consider the problem of estimating a regression function when the distribution of the data belongs to an exponential family. Several interesting problems are under this setting, for example, logit, Poisson, and exponential regressions. We first introduce these three estimation problems and then solve them by means of a new estimation procedure which is called Rho-estimation. Finally, we carry out a simulation study for illustrating and discussing the theoretical properties of rho-estimators as compared to the well-known maximum likelihood ones.

Additional material: