Performance and Security Analysis for GPU-Based Applications
Graphics Processing Units (GPUs) are used today in a number of different
domains, for example in avionics, in cars, and even in health care. However,
these new areas of use bring new requirements for hardware and software in
terms of, for example, performance and security.
In his thesis, Adrian Horga proposes solutions to deal with such requirements.
Best Paper Award BlackboxNLP 2022
Jenny Kunz, Martin Jirénius, Oskar Holmström and Marco Kuhlmann received the
Best Paper Award at the Fifth BlackboxNLP Workshop on Analyzing and
Interpreting Neural Networks for NLP for their paper “Human Ratings Do Not
Reflect Downstream Utility: A Study of Free-Text Explanations for Model
Models able to generate free-text rationales that explain their output have been proposed as an important step towards interpretable NLP for “reasoning” tasks such as natural language inference and commonsense question answering. However, the relative merits of different architectures and types of rationales are not well understood and hard to measure. In this paper, we contribute two insights to this line of research: First, we find that models trained on gold explanations learn to rely on these but, in the case of the more challenging question answering data set we use, fail when given generated explanations at test time. However, additional fine-tuning on generated explanations teaches the model to distinguish between reliable and unreliable information in explanations. Second, we compare explanations by a generation-only model to those generated by a self-rationalizing model and find that, while the former score higher in terms of validity, factual correctness, and similarity to gold explanations, they are not more useful for downstream classification. We observe that the self-rationalizing model is prone to hallucination, which is punished by most metrics but may add useful context for the classification step.
The paper builds on Martin Jirénius’ Master’s Thesis, which was defended earlier this year.
The work on the paper has been partially supported by the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by the Knut and Alice Wallenberg Foundation.
Best conference paper award for 2022 in SWE-CTW 2022
Congratulations to Nikolaos Pappas and his student (who graduated last year
from LiU)! They got the best conference paper award for 2022 in the Swedish
Communication Technologies Workshop (SWE-CTW 2022).
- In this work, we study the interplay of two fundamental notions of timely delivery of information, the deadline-constrained packet delivery and freshness of information. We consider a two-user wireless multiple access setup with random access, in which the first user is equipped with a queue and has external bursty deadline-constrained traffic. The second user monitors a sensor and generates and transmits on-demand status updates. We provide analytical expressions for the throughput and drop probability of the first user, and an analytical expression for the Age of Information (AoI) of the second user. The analysis utilizes discrete-time Markov chains (DTMC). The AoI analysis is not based on the typical graphical approach (sawtooth diagram) but on a more systematic approach utilizing DTMC. Thus, we can provide a complete characterization of the distribution of AoI and the AoI violation probability.
The work has been partially supported by the Swedish Research Council (VR) and by CENIIT.
Prediction Methods for High Dimensional Data with Censored
While there are a wealth of data available in many industrial applications,
these data are often incomplete and strategies are needed to use the data in an
appropriate manner when making predictions. Caroline Svahn contributes with her
doctoral thesis to the field of censored covariates in prediction models by
introducing scalable strategies which can handle complex dependencies in data.
Transforming Service Innovation with Service Design
As services have become increasingly technology-infused, many organizations,
both public and private, are turning to service design as a promising new way
to innovate. Although both research and practice have established that service
design can be an impactful approach to innovate, there is still a dearth of
open questions that need further empirical support. The dissertation titled:
Innovating Innovation: the role of service design in service innovation defined
by Ana Kustrak Korper explores precisely this. How can organizations innovate
their services using service design, and what opportunities and challenges are
they facing? This dissertation articulates two dominant roles that service
design plays in organisational innovation work, and discusses implications for
organisations' value creation and capability building.
From June 6-8, LINHAC, Linköping Hockey Analytics Conference, took place. This
year, LINHAC was an online event broadcasted from Linköping. LINHAC 2022 was
the largest hockey analytics conference in the world and the first in Europe.
The conference gathered top names from academia, hockey clubs, national teams, ice hockey associations, sports analytics industry and entertainment industry from Europe and North America. Further, a student competition was organized with data provided by Sporlogiq and the SHL.
The conference was organized by Patrick Lambrix and Niklas Carlsson from the sports analytics group at IDA together with Mikael Vernblom from Linköping Hockey Club.
Best paper award IFIP Networking 2022
David Hasselquist, Martin Lindblom and Niklas Carlsson received the best paper
award at IFIP Networking 2022 for their paper "Lightweight Fingerprint Attack
and Encrypted Traffic Analysis on News Articles".
In their work, they present a lightweight fingerprint attack and encrypted traffic analysis that allows an attacker to identify what news articles a person is reading on popular news websites even though the traffic is encrypted using HTTPS. They are the first to demonstrate and evaluate the practical effectiveness of this type of attack when applied on modern news websites, and provide valuable insights into how websites can best protect themselves against this type of attacks.
Ontology-Driven Data Access and Data Integration with an Application in the Materials Design Domain
In different fields, not the least in e-science, there is a need for
semantics-aware data access and data integration over heterogeneous databases.
To this aim, in his PhD thesis, Huanyu Li presents a generic ontology-based and
GraphQL-based framework for data access and integration, provides a prototype
implementation and develops an approach for ontology extension, needed when the
databases evolve. Further, he applies and evaluates his approaches within the
materials design domain for which ontologies are created and extended, and a
system is implemented for accessing and integrating data in materials
Bachelor and Master thesis awards
The IDA best thesis award in collaboration with the Computer Society was
announced in an award ceremony at Vallfarten for the 23rd time. The two prize
winning theses were selected among 6 nominated ones that qualified for
selection by this year's Promoter Kristian Sandahl.
Automatic Adaptation of Swedish Texts for Increased Inclusion
To provide texts written in Easy Language is important, as it allows more
people to read and assimilate written text. However, to adapt texts manually is
resource consuming and if this process could be automated, it would be
beneficial both for the individual reader, as well as for society. In her PhD
thesis, Evelina Rennes adopts a target audience perspective and links existing
guidelines for writing Easy Swedish to a theoretical model of reading
comprehension and evaluates them based on applicability. The thesis presents a
palette of methods and techniques for automatically adapting text to Easy
Swedish, as well as techniques for visualizing text complexity and creating
text resources suitable for automatic text adaptation.
Synthesis of Safety-Critical Real-Time Systems
Modern safety-critical real-time systems are becoming more and more complex due
to sophisticated applications. This has led to increased requirements for
communication infrastructures. Time-Sensitive Networking (TSN) is a promising
technique that can handle the increasing bandwidth requirements while meeting
real-time constraints. In his dissertation, Yuanbin Zhou has studied the
synthesis of schedules and routes for TSN, in order to fulfill timing and
reliability requirements for safety-critica
Funding for course development
Within the Vice Chancellor's new investment project to support initiatives
related to course development, Marco Kuhlmann, Annika Tillander and Isak
Hietala have received funding.
SSF funding for IDA
SSF, Swedish Foundation for Strategic Research’s investment in software
development, provides SEK 30 million in funding for the research project
“Adaptive software for the heterogeneous edge-cloud continuum” applied for by
Christoph Kessler with co-applicants Mikael Asplund, Niklas Carlsson, Zebo Peng
and Soheil Samii.
Most influential paper awarded at ICST’22
The most influential paper in the past industry tracks of IEEE International
Conference on Software Testing, Verification and Validation (ICST) was won by a
2012 contribution by Leif Jonsson, David Broman, Kristian Sandahl, and Sigrid
Eldh with the title “Towards Automated Anomaly Report Assignment in Large
Complex Systems using Stacked Generalization".
EU grant for project with focus on circular economy
IDA's Eva Blomqvist (Division of Human-Centered Systems) will lead Onto-DESIDE,
a EU-funded Horizon Europe project on circular economy. The project includes
reuse, remanufacturing and recycling of products and materials in industry with
a focus on ontology-based decentralized sharing of industry data. The project
will run for 3 years and received circa 3.3 MEURO in funding. From IDA's
Division of Database and Information Techniques, Olaf Hartig, Niklas Carlsson
and Patrick Lambrix participate.
WASP Alumni of the year 2021
Olov Andersson, former WASP PhD student from Linköping University, today a WASP
funded postdoc at ETH Zürich, has been awarded WASP Alumn of the year 2021.
This was announced at the yearly WASP Winter Conference in January 2022.
Through the Alumni of the Year award, WASP acknowledges individuals who have excelled within WASP by generating outstanding scientific impact as well as contributions to the program by engagement in the WASP community.
- Olov is a role model for current WASP PhD students and demonstrates the ability to create unique career possibilities and memorable experiences beyond the PhD, says Anders Ynnerman, Director of WASP.
Page responsible: Webmaster