Events - Page 2

Time and place: , Rom Ø434, Fysikkbygningen

Towards Understanding Robustness of Neural Networks using Local Learning Rules

Time and place: , Lille fysiske auditorium, Fysikkbygningen

Analytic description for synchronization of two-level quantum systems

Time and place: , Rom Ø397, Fysikkbygningen

Transport in the one-dimensional spin-S XXZ model

Time and place: , Lille fysiske auditorium, Fysikkbygningen

Analysis of the Functional Role of Directed Simplicial Structures in Biological Neural Networks

Time and place: , Lille fysiske auditorium, Fysikkbygningen

Finite Element and Neural Network Solvers for Modelling Microcirculation

Time and place: , Lille fysiske auditorium, Fysikkbygningen

A Novel Application of Machine Learning to Develop Pointing Models for Current and Future Radio/Sub-millimeter Telescopes

Time and place: , Rom Ø434, Fysikkbygningen

Classification and feature Regression for Multi-Phase Flow Regimes

Time and place: , Center for Computing in Science Education (The kitchen area)

The toolbox of algebraic manipulations that students traditionally learn to use in second-year calculus is not a good match for many applications in other disciplines, such as physics.  Mastery of electromagnetism, for example, requires a geometric understanding of vector fields and their derivatives.

Furthermore, most mathematical modeling requires a robust understanding of the relationship between discrete data and its idealization as smooth mathematical objects.  These applications require students to have rich concept images of differentiation and integration that go well beyond what is typically taught in second-year calculus.

This talk describes efforts at Oregon State University to help students master the use of geometric reasoning in such physical and geometric contexts, in both mathematics and physics courses.  Several examples will be presented where language differences between disciplines lead to student difficulties, as will some of the methods and tools that we have developed to address them.

Time and place: , Celsius

Constrained folding dynamics: a generalized model for labyrinth pattern development

Time and place: , Rom Ø434, Fysikkbygningen

Deployment of semi-unsupervised learning in the search for new physics at the LHC with the ATLAS detector

Time and place: , Center for Computing in Science Education (The kitchen area)

I will show you who are the members of the Physics Education research group of the University of Bologna. Moreover, I will tell you about the projects in which we are involved now and all the past significnt projects that have helped to strengthen the direction and current identity of the group. Also I will indicate what are the main publications that have been made in these years.

All this will be done through the eyes and the narrative of a PhD student part of UNIBO research group. I have not certainly the presumption to know exactly (after two years) how to orient myself in all the rich and multi-dimensional research activities of my group.  But I will try to make you perceive the great commitment, thoughts and enthusiasm that I see every day in the challenging but wonderful research activity in Bologna.

Time and place: , Rom Ø434, Fysikkbygningen

Application of Supervised Machine Learning to the Search for New Physics in ATLAS data

Time and place: , Lille fysiske auditorium, Fysikkbygningen

Predicting Frictional Properties of Graphene Kirigami Using Molecular Dynamics and Neural Networks

Mikkel Jensen: 

Various theoretical models and experimental results propose different governing mechanisms for friction at the nanoscale. We consider a graphene sheet modified with Kirigami-inspired cuts and under the influence of strain. Prior research has demonstrated that this system exhibits out-of-plane buckling, which may cause a decrease in contact area when sliding on a substrate.

According to asperity theory, such a decrease in contact area is expected to reduce friction. However, to the best of our knowledge, no previous studies have investigated the frictional behavior of a nanoscale Kirigami graphene sheet subjected to strain.

William Hirst: 

This thesis explores a diverse array of Machine Learning (ML) models as they search for chargino-neutralino pair production in three-lepton final states with missing transverse momentum. The study is based on a data set of sqrt(s) = 13 TeV proton-proton collisions recorded with the ATLAS detector at the LHC, corresponding to an integrated luminosity of 139 fb−1. The ML models applied in the study were three variants of Deep Neural Networks (DNN), and Boosted Decision Trees (BDT). The DNN variants included an ordinary dense Neural Network (NN), Parameterized Neural Network (PNN) and ensemble models utilizing pattern-specific pathways created by competing neurons. In the latter variant I included a novel layer introduced in this thesis, the Stochastic-Channel-Out (SCO).

Time and place: , Center for Computing in Science Education (The kitchen area)

Fermi problems (also known as Fermi estimation problems or estimation problems) are a classic type of exercise that have been used in physics education for decades.

At UiO we have begun using them in our first-semester physics course as a gentle start to “thinking like a physicist." But Fermi problems can also be a rich site for building modeling skills across the STEM disciplines, and when combined with a little statistics and computation can quickly turn into simple Monte Carlo problems.
 

This ODD seminar will be partly a Fermi problem workshop and partly a demonstration of the new class of computational Fermi problems we are developing for our physics and STEM courses.
 

Also, there will be cake!

We have restructured the first year of the physics bachelor program, introducing a new course combining numerical methods with introductory mechanics in the first semester. I will describe briefly the background for this and then discuss some of the experiences that we have had running the course almost twice by now. I will focus on the types of numerical problems the students have worked on and describe what they have mastered and what they have found challenging.

Anders Lauvland: Physics identity has risen as a much studied construct in recent years, and recognizing oneself as a “physics person” appears to be indicative of persistence in physics. In this work, we integrate a “physics person” construct into expectancy-value theory as a mediator for motivation.
 

For this analysis we have surveyed N=328 first-year students in introductory mechanics at five research intensive institutions in Norway. And use structural equation modelling to analyze the data.

Daniel Johan Aarstein: 

Within fluid mechanics, most interesting phenomena occur on the boundary between fluids of different densities, i.e. water+air, water+oil. Adding the constraints that the two fluids are insoluble, in addition to having the system take place in a pipe, we might experience what is known as a "slug".

Experimental and numerical study of slug behavior is a field within itself, this thesis aims to be a proof-of-concept that a novel, non-intrusive Deep Learning model can be used for real-time analysis. The model itself utilizes a Convolutional Neural Network in order to classify, and predict properties for a given slug in a pipe, based solely on acoustic emission from said pipe.


Current findings indicate that the classification on unseen data has an accuracy of ~93 %. The regression for velocity and length is, however, less precise with R2 scores of ~0.5 and ~0, respectively.

João Inácio: 

The field of low-dimensionality magnetism has developed into an active area of solid-state physics, attracting both theoretical and experimental researchers. Due to the vast array of theoretical tools, there is a large effort to develop a full theoretical understanding of one-dimensional (1D) systems. A large part of this interest is due to nonequilibrium dynamics, where steady-state transport is a generic example.

Many physically relevant 1D models are Bethe-ansatz integrable, such as the spin-1/2 XXZ-model and the Fermi-Hubbard model, but computing transport coefficients still poses a great challenge. Moreover, little is currently known about transport properties of non-integrable models, such as the spin-S XXZ-model or ladder spin systems.

Omid Mirmotahari: For feedback processes to be enhanced, students need both appreciation on how feedback can operate effectively and strategies to use feedback within the curriculum. Several studies that examined feedback exchanged during student peer-review process have shown some challenges regarding students’ confidence in their own knowledge to perform assessment.

It has also been found that, due to students’ negative perception of their own level of competence, they have limited confidence in their fellow students assessing their assignment. Lack of clarity about assessment criteria and standards are a source of anxiety for many first-year university students.

Maria Vetleseter Bøe: We need students to engage in the learning activities that work. However, many students resist active learning activities and believe they learn more from traditional lectures.

We have studied how students’ motivation interacts with different learning activities, using focus group data from Norwegian physics students.

The findings suggest that students develop forms of motivation that promote learning in active learning situations when they feel competent, either through mastery, perceived learning, or comparison with peers, and that relatedness to a community of learners often helps enable development of such motivation.

Sakarias Frette: The standard model is the most accurate theory to date, with incredible precision measurements done at multiple detectors. It has however some shortcommings, not being able to explain phenomena such as the hierarchy problem, gravity, dark matter, etc.. Additional theories have been put forward to try to cover these issues, but for now, it has yielded no luck.

The strategy until recently has been to take such a model, and do a targeted search, resulting in large exclusion plots and no new physics. This is effective and fast for a single model, but very biased, and takes a lot of time if you want to try on 100 og 1000 models.

My thesis will instead try to apply a semi unsupervised technique to separate out anomalous data such that we can reduce the uninteresting SM background and focus on possible new physics that might be hidden in the data. The data analysis model used is an autoencoder.

Ben Zwickl from Rochester Institute of Technology is back an even week:

This is an in-progress project, and I’m looking forward to an interactive discussion with lots of questions and feedback!

Our team has conducted 16 interviews with scientists from chemistry, physics, biology, and mathematics. Each interview delves into important tasks that incorporate computation as part of doing research and/or teaching.

Henning V. Myhrehagen, Karl Henrik Fredly, and Hannah Christine Sabo are working on separate research projects in Tor Ole Odden’s research group at CCSE. Their projects address computation/programming in different parts of physics education. Here are the abstracts of the three lightning talks they will give: