Events

Nov 24 - Nov 30, 2019

Statistical Analysis of Multi-day Solar Irradiance using a Threshold Time Series model

Carolina D. Euan Campos, Postdoctoral Research Fellow, Statistics

Nov 28, 12:00 - 13:00

B9 H2 R2325

solar energy solar irradiance analysis nonlinearity TAR model statistical analysis forecasting

Abstract The analysis of solar irradiance has important applications in predicting solar energy production from solar power plants. Although the sun provides every day more energy than we need, the variability caused by environmental conditions affects electricity production. Most of the existing statistical models to forecast solar irradiance are linear and highly depend on normality assumptions. However, solar irradiance shows strong non-linearity and is only measured during the day time. Thus, we propose a new multi-day threshold autoregressive (TAR) model to quantify the variability of the
Plasmonics in wide-band-gap semiconductor materials

Prof. Jing Li, Pen-Tung Sah Institute of Micro-Nano Science and Technology, Xiamen University

Nov 27, 16:30 - 17:30

B3 L2 R2255

Recently, surface plasmon resonance (SPR) effect has been widely applied in wide-band-gap semiconductor materials (e.g. GaN, ZnO or TiN etc.) for emission enhancement, absorption regulation, sensitive bio- or chemical detections and so on. In this talk, different metal (Ag or Al) nanostructures were introduced into and successfully fabricated on several kinds of wide-band-gap semiconductor materials, including ZnO, AlGaN, TiN, and h-BN by template-based nanosphere lithography method and deposition techniques.
Asymptotically Exact Variational Bayes for High-Dimensional Probit Models

Prof. Daniele Durante, Department of Decision Sciences, Bocconi University, Italy

Nov 27, 15:30 - 16:30

B1 L4 R4102

Bayesian approach to inference decision sciences

Abstract There are several Bayesian models where the posterior density is not available in a closed and tractable form. In these situations, Markov chain Monte Carlo algorithms and approximate methods, such as variational Bayes and expectation-propagation, provide common solutions to perform posterior inference. However, in high-dimensional studies, with large or even huge p, these approaches still face open problems in terms of scalability and quality of the posterior approximation. Notably, such issues also arise in basic predictor-dependent models for binary data which appear in a wide
Hidden Backdoors in Deep Learning Systems

Prof. Ben Zhao, Computer Science, University of Chicago, USA

Nov 25, 12:00 - 13:00

B9 L2 H1 R2322

In this talk, I will describe two recent results on detecting and understanding backdoor attacks on deep learning systems. I will first present Neural Cleanse (IEEE S&P 2019), the first robust tool to detect a wide range of backdoors in deep learning models. We use the idea of perturbation distances between classification labels to detect when a backdoor trigger has created shortcuts to misclassification to a particular label. Second, I will also summarize our new work on Latent Backdoors (CCS 2019), a stronger type of backdoor attack that is more difficult to detect and survives retraining in commonly used transfer learning systems. Latent backdoors are robust and stealthy, even against the latest detection tools (including neural cleanse).
Planning 5G under EMF constraints: Challenges and opportunities

Prof. Luca Chiaraviglio is a Tenure Track Assistant Professor at the University of Rome Tor Vergata (Italy)

Nov 24, 13:30 - 14:30

B1 L3 R3119

5G

Abstract 5G networks are currently facing the first installation steps in many countries. From a user perspective, there is a growing concern about the health impact of the ElectroMagnetic Fields (EMFs) generated by 5G. in this context, different countries in the world adopt EMF limits defined by the International Commission on Non-Ionizing Radiation Protection (ICNIRP), but many countries, including Italy, adopt much more stringent regulations, by following precautionary principles. In this talk, we will focus on the planning of 5G networks under strict EMF limits. Starting from a set of
Network Function Virtualization: What, Why and How?

Ahmed E. Kamal, Department of Electrical and Computer Engineering, Iowa State University, Ames, IA, USA.

Nov 24, 11:00 - 12:00

B1 L3 R3119

Virtualized Network Function

The European Telecom Standards Institute (ETSI) introduced the concept of Network Function Virtualization (NFV) with the aim of efficient network architecture and network system operation. In traditional networks, network functions are implemented in dedicated physical machines which are designed for single functionalities. Network services have been provided by connecting these physical machines, so the network architecture has been highly rigid and hard to change. NFV environment provides a more flexible and scalable network configuration and implementation through the softwarization of physical network functions. Network functions are transformed to Virtual Machines (VMs) so that Virtualized Network Functions (VNFs) can be implemented in commodity servers built for common uses, including public clouds.

Nov 17 - Nov 23, 2019

Nov 10 - Nov 16, 2019

Nov 3 - Nov 9, 2019

Oct 27 - Nov 2, 2019

Energy-conserving and energy-stable discretizations of differential equations

David Ketcheson, Professor, Applied Mathematics and Computational Sciences

Oct 31, 12:00 - 13:00

B9 L2 H1 R2322

numerical discretization energy stability energy conservation Runge-Kutta methods

Abstract Many physical models are characterized by the property that some measure of energy is conserved or is non-increasing in time. This property may be challenging to guarantee in numerical discretizations; often it requires the use of costly implicit methods. I will review some approaches to energy stability and describe a new approach that guarantees energy stability with a general class of explicit schemes that are a simple modification of standard Runge-Kutta methods. Brief Biography David Ketcheson is an Associate Professor of Applied Mathematics and Computational Science (AMCS) and

Oct 20 - Oct 26, 2019

Oct 13 - Oct 19, 2019

Oct 6 - Oct 12, 2019

Sep 29 - Oct 5, 2019

Task-driven Automated Data Visualization

Dr. Yunhai Wang, Professor, Computer Science, Shandong University, China

Oct 2, 12:00 - 13:00

B9 L2 H2 R2325

data visualisation visualisation techniques parameters time series data high dimensional data EdWordle ShapeWordle

Abstract By providing visual representations of data, visualization can help people to carry out many data analysis tasks more effectively. Given a data set, however, there are often too many possible visualization techniques, with each technique having many parameters to be tuned. This arises the question is if it is possible to automatically design a visualization that is best suited to pursue a given task on given input data. With my group, I developed a number of techniques to achieve this goal for different data sets such as time-series data, high-dimensional data, text as well as graphs
BGP hijacks: the next generation stealthy Command and Control for botnets

Dr. Marc Dacier, Chair of the Digital Security department and a full Professor at Eurecom, France

Sep 30, 12:00 - 13:00

B9 L2 H1 R2322

botnets C2C malware command and control

Abstract It is well known that malware spreading over the Internet aim at transforming vulnerable devices into bots that can be misused by attackers. These armies of bots constitute what is commonly called "botnets" and they are given tasks (such as spamming, dosing, etc..) to do through a "command and control" infrastructure (C2C). Identifying and neutralizing these C2C has been the subject of an arms race between white and black hats for years. In this talk, we will briefly explain how C2C works and how they have been (and still are being) detected. We will then present some very strange