Elektro- och informationsteknik

Lunds Tekniska Högskola | Lunds universitet


SoS Workshop 2019

Publicerad: 2019-04-08

Save the date

När: 2019-09-19 09:00 till 2019-09-20 15:00
Plats: Grand Hotel dag 1, LTH dag 2


AIML@LU WS: AI & ML Technologies

Publicerad: 2019-03-20

This AIML@LU fika-to-fika workshop focuses on the development of the technologies that form the basis of Artificial Intelligence and Machine Learning. Possible topics to discuss are the research front for different types of AI, but also to look at different techniques for machine learning.

Room E:A, E-huset, Ole Römers väg 3 LundWhen: 30 August at 9.30 - 15.30  

Where: E:A, E-building, Ole Römers väg 3, LTH, Lund University


9.30 Fika and mingle

10.15 Introduction and update regarding the AIML@LU network

10.30 Ongoing projects

Collaborative reading robotMartin Karlsson, Lund University: Robot Programming by Demonstration Based on Machine Learning

Abstract: Whereas humans would prefer to program on a high level of abstraction, for instance through natural language, robots require very detailed instructions, for instance time series of desired joint torques. In this research, we aim to meet the robots half way, by enabling programming by demonstration.

Marcus Klang, Lund University:  Finding Things in Strings

Najmeh Abiri, Lund University: Variational Autoencoders

Joakim Johnander, Linköping University: Deep Recurrent Neural Networks for Video Object Segmentation

12.00 Lunch, mingle and poster session

13.00 Future trends and interesting examples

Mikael GreebMikael Green, Desupervised2: Bayesian Deep Probabilistic Programming: Are we there yet?

Abstract: Not many would argue against the Bayesian paradigm being the most useful one in modeling problems where parameter estimations are inherently uncertain. But unfortunately most interesting models, especially the ones we know from deep learning, have been very hard to fit in any reasonable amount of time. When dealing with +10 million parameters and +100 thousand data points, Markov Chain Monte Carlo just isn't a viable option. This is why almost every practitioner in deep learning defaults to maximum likelihood estimates through optimization via stochastic gradient descent, because it's much faster. In this talk we'll explore a promising way of doing full Bayesian inference on large scale models via stochastic black box variational inference.

ErikErik Gärtner, Lund University: Intrinsic Motivation - Exploration, curiosity and learning for learning's sake

Abstract: Humans as well as other animals are curious beings that develop cognitive skills on their own without the need for external goals or supervision.
Inspired by this, how can we encourage AIs to learn and solve tasks by themselves?
This talk presents the fascinating area of intrinsic reward in the context of reinforcement learning by showcasing recent articles and results.

14.30 Summary and conclusions

15.00 Fika and mingel



To participate is free of charge, but please register no later than 28 August 12.00 at:


If you have any questions, suggestions or would like to contribute to the program please contact one of:

More AIML@LU events at | Join the AIML@LU Network at:

När: 2019-08-30 09:30 till 2019-08-30 15:30
Plats: E:A, E-building, Ole Römers väg 3, LTH, Lund University


ELLIIT Distinguished Lecture by Bill Dally: The Future of Computing: Domain-Specific Accelerators

Publicerad: 2019-06-17

Speaker: Bill Dally, Chief Scientist and Vice President of NVIDIA Research and Inez Kerr Bell Professor of Computer Science and Electrical Engineering at Stanford University

Title of talk:  The Future of Computing:  Domain-Specific Accelerators

Abstract: Scaling of computing performance enables new applications and greater value from computing. With the end of Moore?s Law and Dennard Scaling, continued performance scaling will come primarily from specialization. Graphics processing units are an ideal platform on which to build domain-specific accelerators. They provide very efficient, high performance communication and memory subsystems - which are needed by all domains. Specialization is provided via ?cores?, such as tensor cores or ray-tracing cores that accelerate specific applications. This talk will describe some common characteristics of domain-specific accelerators via case studies.

Room: MH:309A

When: Wednesday June 19, 14:30-15:00

När: 2019-06-19 14:30 till 2019-06-19 15:00
Plats: MH:309A, Mattehuset, Sölvegatan 18, Lund


AI* Nordic Powwow

Publicerad: 2019-03-18

Exploring AI ? The future benefits and challenges. Beyond the traditional conference we present the unique POWWOW experience.

During this POWWOW we gather across sectors and competences around the hot topic Artificial Intelligence. Together we explore the impact it will have on society and our common future, but we also look into specific industries and their applications.

One of the purposes of the day will be the meeting between people and building an active platform for collaboration and innovation moving forward. All while having fun.

Some speakers:

  • ANDERS BORG, AI Adviser IPsoft, Fintech investor and Former Minister of Finance in Swedish Government

  • JOSÉ VAN DIJCK, Professor in media and digital society at Utrecht University

  • LIJO GEORGE, Business Lead at Sony AI

  • LISELOTT LADING, Serial entrepreneur, board member and Business Integrator IT, Axis Communications AB

  • KALLE ÅSTRÖM, Professor in Mathematics at Lund University

  • DAVID POLFELDT, CEO Massive Entertainment AB

More speakers, deatils of the program and how to register at the conference website: 


Conference fee: SEK 1.950 (Fee excluding VAT). Employees at Lund university are given a discount of 25%. Contact cecilia [at] skanemotor [dot] se (Cecilia Löfberg) at cecilia [at] skanemotor [dot] se or +46 707 88 40 48 for discount code.


Nordic Artificial Intelligence Powwow  is a collaboration between Skånemotor and Lund University.

More AIML@LU events at:

När: 2019-05-23 09:00 till 2019-05-23 22:00
Plats: Central Lund, Sweden


Massive MIMO: Prototyping, Proof-of-Concept and Implementation - PhD Defence by Steffen Malkowsky

Publicerad: 2019-04-23

Author: Steffen Malkowsky, Department of EIT

Location:  E:1406, E-building, Ole Römers väg 3, LTH, Lund University

Faculty opponent:  Professor Joseph R. Cavallaro

Thesis for download (PDF) 


Wireless communication is evolving rapidly with ever more connected devices and significantly increasing data rates. Since the invention of the smartphone and the mass introduction of mobile apps, users demand more and more traffic to stream music, watch high-definition video or to simply browse the internet. This tremendous growth is more pronounced by the introduction of the Internet of Things (IoT) in which small devices, such as sensors, are interconnected to exchange data for all sorts of applications. One example are smart homes in which a user can for instance, check temperature at home, verify if windows are closed or open, or simply turn on and off distributed loud speakers or even light bulbs in order to fake a busy household when on vacation. With all these additional devices demanding connectivity and data rates current standards such as 4G are getting to their limits. From the beginning 5G was developed in order to tackle these challenges by offering higher data rates, better coverage as well as higher energy and spectral efficiencies. Massive Multiple-Input Multiple-Output (MIMO) is a technology offering the benefits to overcome these challenges. By scaling up the number of antennas at the Base Station (BS) side by the order of hundred or more it allows separation of signals from User Equipments (UEs) not only in time and frequency but also in space. Exploiting the high spatial degrees-of-freedom it can focus energy with spotlight precision to the intended UE, thereby not only achieving higher energy being received per UE but also lowering the interference among different UEs. Utilizing this precision, massive MIMO may serve a multitude of UEs within the same time and frequency resource, thereby achieving both higher data rates and spectral efficiency. This is a very important feature as spectrum is very crowded and does not allow for much higher band-widths, and more importantly is also very expensive. 

The promised gains, however, do come at a cost. Due to the significantly increased number of BS antennas, signal processing and data distribution at the BS become a challenging task. Signal processing complexity scales with the number of antennas, thus requiring to distribute different tasks properly to still achieve low-latency and energy efficient implementations. The same holds for data movement among different antennas and central processing units. Processing blocks have to be distributed in a manner to not exceed hardware limits, especially at points where many antennas do get combined to perform some kind of centralized processing. 

The focus of this thesis can be divided into three different aspects, first, building a real-time prototype for massive MIMO, second, conducting measurement campaigns in order to verify theoretically promised gains, and third, implementing a fully programmable and flexible hardware platform to efficiently run software defined massive MIMO algorithms. In order to construct a prototype, challenges such as low-latency signal processing for huge matrix sizes as well as task distribution to lower pressure on the interconnection network are considered and implemented. By partitioning the overall system cleverly, it is possible to implement the system fully based on Commercial off-the-shelf (COTS) Hardware (HW). The working testbed was utilized in several measurement campaigns to prove the benefits of massive MIMO, such as increased spectral efficiency, channel hardening and improved resilience to power variations. Finally, a fully programmable Application-Specific Instruction Processor (ASIP) was designed. Extended with a systolic array this programmable platform shows high performance, when mapping a massive MIMO detection problem utilizing various algorithms, while post-synthesis results still suggest a relatively low-power consumption. Having the capability to be programmed with a high-level language as C, the design is flexible enough to adapt to upcoming changes in the recently released 5G standard.

När: 2019-05-17 09:15 till 2019-05-17 09:15
Plats: E:1406, E-building, Ole Römers väg 3, LTH, Lund University


5G - An Antenna and Measurements Perspective

Publicerad: 2019-01-25

An exciting event oreganised by The Antenna Measurement Techniques Association. Spend the day with the Antenna Measurement Techniques Association listening to top experts present the most recent developments in the industry.

Technical Tour

May 6, 2019, 18:00?20:00

Arrive a day early so you can plan to join us as we take a tour of the MAX IV Laboratory. Transportation and tour are included in the price of registration.
Technical Program

Programme May 7, 08:00?19:00

DTU-ESA Spherical Near-Field Antenna Test Facility ? Past, Present, and Future Activities
by Prof. Olav Breinbjerg, Technical University of Denmark, Lyngby, Denmark

Near-Field Measurement Technique for Electromagnetic Exposure of 5G Devices
by Prof. Mats Gustafsson, Lund University, Sweden

Far-Field OTA Testing of User Equipment Using Plane Wave Generators
by Mr. Lars Foged, Microwave Vision Group (MVG), Italy

5G Over-The-Air Conformance Testing
by Dr. Jonas Fridén, Ericsson Research, Gothenburg, Sweden

5G: Challenges for Human Exposure Assessment and Virtual-Drive Over-the-Air Testing
by Dr. Christian Bornkessel, Technische Universität Ilmenau, Germany

High-Resolution Dynamic Characterization of mm-Wave Channels
by Prof. Fredrik Tufvesson, Lund University, Sweden

Organizing Committee

Christer Larsson
Donnie Gray
Manuel Sierra Castañer
Michael Havrilla
Michelle Taylor
Fredrik Tufvesson
Lars Foged
Jan Zackrisson

Registrationis now open at:

The full program and information about fees (PDF)


När: 2019-05-07 08:00 till 2019-05-07 19:00
Plats: Lund University Student Union (Kårhuset) LTH John Ericssons väg 3, Lund Sweden


Sidansvarig: |