DISA

Centre for Data Intensive Sciences and Applications

Welcome to our November PhD-seminar

2023-10-12

When? Friday November 3rd 14-16
Where? Onsite: D1140 at Linnaeus University in Växjö and online
Registration: Please sign up for the PhD-seminar via this link https://forms.gle/txd4g2Un1qfYSvr2A by November 1st (especially important if you plan on attending onsite so we have fika for everyone)

Agenda
14.00-14.10 Welcome and practical information from Welf Löwe
14.10-14.55 Presentation and discussion: Unlock the Power of Mobile Robotics with AI based Vision – Daniel Nilsson, PhD student at Kuka Nordic
14.55 – 15.05 Coffee break
15.05 – 15.50 Presentation and discussion – AI and data literacy: What knowledge and skills are needed in a data-driven society? – Johanna Velander, PhD student UPGRADE and WASP-HS
15.50 -16.00 Sum up and the Big Data Conference and plan for our next seminar on January 12th

Abstracts
Unlock the Power of Mobile Robotics with AI based Vision – Daniel Nilsson, PhD student at Kuka Nordic
This research project proposes a transformation of KUKA’s AMRs into multifunctional agents using the vehicles existing sensors and AI. In the current solutions Autonomous Mobile Robots (AMRs) primarily serve as single-function devices, enabling driverless transportation between various points within a facility. At the same time automation of processes such inventory tracking tend to involve additional investments in additional systems and costly hardware installations, adding to the financial dilemma when striving to transform the operation toward industry 4.0. By allowing AMRs to manage transportation, inventory tracking, and safety compliance simultaneously, such advancement would significantly add to the AMR’s value in production environments.

AI and data literacy: What knowledge and skills are needed in a data-driven society? Johanna Velander, PhD student UPGRADE and WASP-HS
Uncovering patterns and trends in vast, ever-increasing quantities of data has been enabled by different machine learning methods and techniques used in many of the applications that we use in our daily lives. Permeating many aspects of our lives and influencing our choices, development in this field continues to advance and increasingly impacts us as individuals and our society. The risks and unintended effects such as bias from input data or algorithm design have recently stirred discourse about how to inform and teach about AI in K-12 education. As AI is a new topic not only for pupils in K-12 but also for teachers, new skill sets are required that enable critical engagement with AI.

In this presentation, I will talk about my PhD project, which sits at the intersection of computer science and teacher education. In a recent study deploying a Learning Analytics plugin at some LNU courses students’ thoughts, attitudes and emotions were investigated when engaging with their own data (collected by the LMS Moodle). Results revealed a low awareness of data collection and potential data-driven practices and also worries about how this data could be used and who could have access to it. Following these insights and according to my Ph.D. affiliation with UPGRADE and WASP-HS I have continued to investigate how awareness and knowledge of AI concepts, applications and potential ethical concerns often referred to as AI literacy can be taught at a K-12 level in order to inspire future data scientists and to enable equal participation in a digital data-driven society according to a critical literacy perspective that empowers learners to act on and find alternatives to issues present in current AI practice.

 

Workshop (May 4th, 2023) on Self-supervised deep learning in EO-based forest inventory – ESA RepreSent project and Forest Thematic Exploitation Platform (F-TEP)

2023-04-27

Deep learning (DL) and computer vision are rapidly gaining popularity in forest inventory. However, the scarcity of available reference data limits the effective use of DL tools. Self-supervised learning (SSL) and weakly-supervised learning aim to solve this bottleneck by enabling better utilization of available EO data to effectively train DL models.

We invite you to attend an online workshop, where we will present and discuss several deep learning models suitable for forest mapping with satellite remote sensing data, that were created within the ESA funded RepreSent project (2022-2023) on Representation learning for Copernicus Sentinel data. The developed models enable forest mapping and monitoring by significantly reducing the amount of reference data typically required for deep learning model training. A selected set of tools has also been implemented on Forestry TEP to facilitate the quick adoption of developed methodologies in the downstream sector and for potential use as benchmark methodologies.

The workshop targets AI4EO researchers who are interested in the forestry sector, as well as foresters who wish to explore the broader applications of DL and SSL in their academic research or operational forest management.

Please inform about your participation using this link https://forms.office.com/e/GFKbeZ29jQ

Participation to the workshop is free.

The online workshop will be organized on MS Teams, and further details will be sent to registered participants.

Preliminary agenda – the online workshop starts at 10 am EEST (Finland time zone), May 4th 2023

10:00 am (EEST) Welcome and ESA RepreSent project introduction Matthieu Molinier, Oleg Antropov, VTT, Corneliu Octavian Dumitru, DLR
10:05 Forest inventory using EO data Jukka Miettinen, Tuomas Häme, VTT
10:15 Self-supervised and weakly-supervised Learning in Earth Observation ESA Represent consortium
10:25

10min + 5min Q&A

 MoCo & MAML models in forest mapping using Copernicus Sentinel-2 and Sentinel-1 data Lloyd Hughes, Marc Russwurm,

Devis Tuia, EPFL
10:40

10min +5 min Q&A

 UNet+ models with multi-source EO data Oleg Antropov, VTT
10:55

10 min +5 min Q&A

 DCVA approaches for forest change detection using Sentinel-2 images Ridvan Kuzu, DLR
11:10 Break (5 mn)
11:15

20 min +5 min Q&A

 F-TEP introduction: Status and tools overview Jukka Miettinen, Renne Tergujeff, VTT
11:40

20 min + 5 min Q&A

 F-TEP service demonstrations including SSL Lauri Seitsonen, VTT
12:05 F-TEP developer’s perspective Lauri Seitsonen, VTT
12:15 Concluding remarks Oleg Antropov, Matthieu Molinier, VTT

Please forward this invitation to your colleagues who might be interested in these topics.

We look forward to meeting you at the workshop.

Oleg Antropov, Matthieu Molinier, VTT and the ESA RepreSent team

Welcome to our May PhD-seminar in 2023

2023-04-12

  • When? May 5th 14.00 – 16.00
  • Where? D1172 – Växjö (link will provided for those who wants to attend online)
  • Registration: We would like to know how many that will attend onsite/online in order to get some fika for those onsite. So please register by May 3rd https://forms.gle/R1GuWXGXjiDYGWaQ6

Agenda
14.00-14.10 Welcome and practical information from Welf Löwe
14.10-14.55 Presentation and discussion: Title Data intensive applications and service development at Volvo CE – Joel Cramsky, Industry PhD-student Volvo CE
14.55 – 15.05 Coffee break
15.05 – 15.50 Presentation and discussion – Designing an Intelligent Predictive Maintenance Framework for Cyber-Physical Systems using Machine Learning and Digital Twin Technology – Mehdi Saman Azari, PhD student at LNU
15.50 -16.00 Sum up and plan for our next seminar on June 2nd

Abstracts

Läs resten av detta inlägg»

Welcome to the April PhD-seminar in 2023

2023-03-09

  • When? April 14th 14.00 – 16.00
  • Where? D1172 – Växjö (link will provided for those who wants to attend online)
  • Registration: We would like to know how many that will attend onsite/online in order to get some fika for those onsite. So please register by April 12th https://forms.gle/JzVPp5h9Uz1Cwaqx6

Agenda
14.00-14.10 Welcome and practical information from Welf Löwe
14.10-14.55 Presentation and discussion: Advanced identification methods for the forest industry through CV/AI – Dag Björnberg, Industry PhD-student Softwerk
14.55 – 15.05 Coffee break
15.05 – 15.50 Presentation and discussion – Sound, Precise, Memory Efficient Points-to Analysis – Mathias Hedenborg
15.50 -16.00 Sum up and plan for our next seminar on May 5th

Abstracts

Läs resten av detta inlägg»

Welcome to the March PhD-seminar 2023

2023-02-08

When? Friday March 10th 14.00-15.00
Where? Via zoom
Registration? No registration needed since the seminar is only online this time, if you have not received the link please contact Diana Unander diana.unander@lnu.se

This time we will only have one presentation and the seminar will be fully online since the presenter is located abroad.

Agenda:
14.00 – 14.10 Welcome and practical information from Welf Löwe
14.10 – 14.55 Presentation and discussion: Using multiple embeddings for visual analytics – Daniel Witschard  (ISOVIS)
14.55 – 15.00 Wrap up – information about the next PhD-seminar

Abstract

Using multiple embeddings for visual analytics – Daniel Witschard  (ISOVIS)

Embeddings are numeric vector representations of complex or unstructured data. The main goal of embedding algorithms is usually to produce embeddings where items that are similar in the original data set are embedded into vectors that lie close to each other in the embedding space. This makes embeddings highly suitable as input for computational analysis tasks such as clustering, classification, and similarity calculations since it is often more straightforward to perform these calculations on numeric vectors than rather than on the underlying data. For some data types, such as graphs/networks and words/text, there exist several different algorithms (each with its specific characteristics and tradeoffs) and therefore choosing the best embedding technology for a given application is an important and often non-trivial task. However, searching for single candidates is not the only strategy that could be used–and therefore this presentation will contain examples and results (some published, and some work-in-progress) aiming to answer the research question “Is it possible to combine several different embeddings to obtain even better results and visualizations?”

More information about Daniels research project https://lnu.se/en/research/research-projects/doctoral-project-multivariate-network-embedding-for-visual-analytics/ 

 

Webinarium om hälsodata och EU-projektet Health Data Sweden

2023-02-02

  • När? 27 feb 2023 kl 12-13
  • Var? Online

Forskargruppen eHälsa och hälsodata i samverkan vid Uppsala universitet bjuder in till webinarium om hälsodata. På progammet finns bland annat:

– Vad innebär egentligen hälsodata och varför är det viktigt?
– Introduktion av EU-projektet Health Data Sweden (HDS)
– Sebastiaan Meijer (KTH) koordinator för HDS
– Maria Hägglund (UU) ansvarig för Uppsalas aktiviteter i HDS
– Stort utrymme för frågor och diskussion

Anmäl dig här  så kommer mer information ungefär en vecka innan webinariet.
Vi ses!

För frågor kontakta: Maria Hägglund (maria.hagglund@kbh.uu.se) eller Sara Riggare (sara.riggare@kbh.uu.se), Forskargruppen eHälsa och hälsodata i samverkan

 

Welcome to the February PhD-seminar 2023

2023-01-18

When? Friday February 3rd 14.00-16.00
Where? D1172 or via zoom, the link will be sent out to those who register
Registration? Register via this link https://forms.gle/t1D3iHkC6mocidQF6 no later than February 1st.

Agenda:
14.00 – 14.10 Welcome and practical information from Welf Löwe
14.10 – 14.55 Presentation and discussion: Digital twin development at Volvo CE (VCE) for Wheel loaders (WLO) –Manoranjan Kumar, Volvo
14.55 – 15.05 Coffee break
15.05 – 15.50 Presentation and discussion – Design and Analysis of Self-Protection: Adaptive Security for Software-Intensive Systems – Charilos Skandylas; LNU
15.50 – 16.00 Sum up and plan for our next seminar on March 3rd

Abstracts

Digital twin development at Volvo CE (VCE) for Wheel loaders (WLO) –Manoranjan Kumar, Volvo CE
In recent years, numerous advancements have been made in technology related to IOT, data visualization, and simulation. Therefore, VCE has decided to develop the digital twin model of Wheel loaders to support and understand the customers need. This decision comes with opportunities and challenges. The success of digital twin depends on a minimum of data being extracted from real machines allowing still estimates of complete vehicle usages using the simulations. Those simulations can be data driven or physics based. VCE’s Digital twin platform supports the developments of hardware as well as software for Digital twins (DT) including:

  • WLO usage DT: Logs of WLO capture the behaviors of drivers and what kind of materials they are handling and classifies the driving styles using machine learning.
  • WLO DT: Complete vehicle simulations of WLOs using the physical properties of different components. The simulation also captures the different controls and path followings. The simulation is “twinned” with the loads measured in real machines. Good correlations are observed for different kinds of driving behaviors.
  • EGR DT: Exhaust gar recirculation (EGR) clogging can be monitored in different types of construction machines using online and offline machine learning approaches.

More information about the research project: https://lnu.se/en/research/research-projects/doctoral-project-digital-twin-developments-within-volvo-ce/

Design and Analysis of Self-Protection: Adaptive Security for Software-Intensive Systems – Charilos Skandylas, LNU
Today’s software landscape features a high degree of complexity, frequent changes in requirements and stakeholder goals, and uncertainty. Therefore, in the corresponding threat landscape cybersecurity attacks are a common occurrence, and their consequences are often severe. Self-adaptive systems have been proposed to mitigate the complexity and frequent degree of change by adapting at run-time to deal with situations not known at design time. They, however, are not immune to attacks, as they themselves suffer from high degrees of complexity and uncertainty. Therefore, suitable software systems that can dynamically defend themselves from adversaries are required. Such systems are called self-protecting systems and aim to identify, analyze and mitigate threats autonomously.

This presentation will discuss approaches with the goal of providing software systems with self-protection capabilities in two parts. The first part aims to enhance the security of architecture-based self-adaptive systems and equip them with self-protection capabilities. Both proactive and reactive self-protection techniques will be discussed. Proactive techniques aim to protect a software system by accurately analyzing its current and future security relevant behaviour and steering the system towards the most secure behavior, minimizing the attack surface. Reactive techniques provide self-protection to an architecture based self-adaptive system via effective countermeasure selection at runtime. In the second part, we extend a classical decentralized information flow control model by incorporating trust and adding adaptation capabilities that allow a full decentralized, open system system to identify security threats and self-organize to maximize the average trust between the system entities while maintaining the security policies of each of the system’s entities.

Invitation: Research Seminar 9/12 11-12

2022-11-15

Title: “Roadmaps for AI Integration in the Rail Sector: Current Project Results and Overview of Case-Studies”

Abstract: Artificial Intelligence (AI) is increasingly affirming as a game-changer technology in several sectors, including rail transport. The overall objective of the H2020 Shift2Rail project RAILS (Roadmaps for AI Integration in the raiL Sector) is to investigate the potential of AI in the rail sector and to contribute to the definition of roadmaps for future research in the context of railway maintenance and inspection, autonomous train driving, and traffic planning and management. This seminar will provide a high-level overview of the RAILS project, presenting the main topics, objectives, ongoing research activities, and preliminary results achieved. Particular attention will be given to the current investigations towards the application of Deep Learning approaches to improve the maintainability of railway assets and the safety of autonomous trains. To be specific, two main case studies will be discussed, and recent advancements presented, concerning smart maintenance at level crossings and vision-based obstacle detection on rail tracks.

Speaker: Lorenzo De Donato (Visiting PhD at LNU) he is a Ph.D. Student in Information Technology and Electrical Engineering. When Lorenzo is not in Sweden visiting us he can be found at the Department of Electrical Engineering and Information Technology, University of Naples Federico II, Naples, Italy

Contact information: lorenzo.dedonato@unina.it

 

Welcome to our first PhD-seminar November 4th

2022-10-15

  • When? November 4th 14-16
  • Where? D1140 – Växjö (link will provided for those who wants to attend online)
  • Registration: We would like to know how many that will attend onsite/online in order to get some fika for those onsite. So please register by November 2nd https://forms.gle/ZwwgoQ4JK4e41BBR6

 Agenda

14.00-14.10     Welcome and practical information from Welf Löwe

14.10-14.55     Presentation and discussion: Visual Analytics for Explainable Machine Learning in a Nutshell – Angelos Chatzimparmpas

14.55 – 15.05  Coffee break

15.05 – 15.50  Presentation and discussion: Getting the most out of health data, combing the best of two worlds – Olle Björneld

15.50 -16.00    Sum up and plan for our next seminar on January 13th

Abstracts

Visual Analytics for Explainable Machine Learning in a Nutshell – Angelos Chatzimparmpas

Machine learning (ML) research has recently gained much attention, with various models proposed to understand and predict patterns and trends in data originating from various domains. Unfortunately, users find it harder to evaluate and trust the results of these models as they become more complex because most of their internal workings are kept in secret black boxes.

One possible solution to this problem is the explanation of ML models with visual analytics (VA) since it enables human experts to analyze large and complex information spaces such as data and model spaces. By doing so, evidence has shown an improvement in predictions and an increase in the reliability of the results.

This talk aims to provide an overview of the state-of-the-art in explainable and trustworthy ML with the use of visualizations, as well as the development of VA systems for each stage of a typical ML pipeline. Furthermore, we will briefly introduce some of these tools and discuss how such VA techniques can help us not only understand ML models but also do this in a human-centered and steerable way.

Getting the most out of health data, combing the best of two worlds – Olle Björneld

Machine learning driven knowledge discovery on real world data based on domain knowledge. Real world data does not comply with machine learning models very well and prediction models perform suboptimal if pre-processing of data is deficient.

Based on experience from medical registry studies using electronic health data (EHR) performed in collaboration with domain experts, data analyst and statistician an automatic feature engineering framework and method have been developed. The framework is called automatic Knowledge Driven Feature Engineering (aKDFE) and have been evaluated by machine learning pipeline.

Experiment shows that prediction models performs better if aKDFE is used without losing explainability, but more experiments need to be performed in other domains to fully quantify the results. The key aspect is how to concentrate and mine inherent knowledge in transaction data to optimal machine learning driven prediction models.

A warm welcome,

Welf & Diana

Webinar on Linked Open Data in Cultural Heritage

2022-02-15

Welcome to the webinar titled: Semantic metadata enrichment and data augmentation of small museum collections following the FAIR principles

Abstract: The lecture will present the benefits and challenges of making the cultural heritage data of small regional museums findable, accessible, interoperable and reusable (FAIR). Using the Archaeological Museum of Tripoli, Greece, as a case study, it demonstrates how the employment of semantic methods, such as semantic enrichment and linking to Linked Data resources, and semantic technologies, such as the CIDOC-Conceptual Reference Model (CRM) and other standard ontologies, can help alleviate some of these challenges and help small museums make their data FAIR. It also discusses how a semantics-based approach can facilitate collaboration between Digital Humanities and Information Studies researchers, and cultural heritage institutions, by providing a common means of communication that means cultural heritage data can be reused, repurposed, and redeployed efficiently.

Welcome!