Thorir Mar Ingolfsson

PhD Student in Robust and Practical Machine Learning

ETH Zurich

Biography

Hi! I am Thorir, and I am currently a Ph.D. student at ETH Zurich under the supervision of Prof. Dr. Luca Benini. My research interest is applying Robust and Practical Machine Learning approaches, focusing on a bio-signal analysis taking into account the characteristics and constraints of wearable edge devices and IoT units. A use case that I am currently researching is the usage of bio-signals to detect and forecast seizures, where the bio-signals come from wearable devices and are classified with ML and DL algorithms on the same wearable edge devices.

Interests

Artificial Intelligence
Neural Architecture Search
Bio-signal Classification

Education

Ph.D. in Electrical Engineering and Information Technology

ETH Zurich
M.Sc. in Electrical Engineering and Information Technology, 2020

ETH Zurich
B.Sc. in Electrical and Computer Engineering, 2018

University of Iceland

Projects

High-Precision Control and Localisation for Robotic Billiard Shots

Taking successful shots on a snooker table is an hard task and requires high-precision control and localisation of the cue. We implemented a vision-based strategy that combines the fields of vision of two cameras, an overhead camera pointing down on the table and a cue mounted camera, allowing a robot arm to obtain the excellent cueing accuracy observed in human players. The robot arm holds a linear motor that plays the role of the cue and the cue camera is mounted on top of the linear motor. Cue pose estimation incorporates orthogonal uncertainty Procrustes analysis. From pose estimation and image information we make feedback corrections that optimally line up the cue so it strikes the cue ball in the desired shot angle. The feedback is then fed to a robot arm that makes the corrections. This method is then implemented and tested on a snooker table with shots of varying difficulty.

Recent & Upcoming Talks

Energy-Efficient Tree-Based EEG Artifact Detection

Energy-Efficient Artifact Detection Based on a Minimal Number of EEG Channels.

14, Jul, 2022 Scottish Event Campus

Energy-Efficient Tree-Based EEG Artifact Detection

Reducing Neural Architecture Search Spaces with Training-Free Statistics and Computational Graph Clustering

Hardware-Aware Neural Network Analysis using Training Free Aspects.

17, May, 2022 Castle of Valentino, Turin, Italy

Reducing Neural Architecture Search Spaces with Training-Free Statistics and Computational Graph Clustering

Hardware-Aware Neural Network Analysis

Hardware-Aware Neural Network Analysis using Training Free Aspects.

15, Oct, 2021 StageOne Event & Convention Hall Zurich

Featured Publications

Thorir Mar Ingolfsson, Simone Benatti, Xiaying Wang, Adriano Bernini, Pauline Ducouret, Philippe Ryvlin, Sandor Beniczky, Luca Benini, Andrea Cossettini

February 2024 In Nature

Minimizing artifact-induced false-alarms for seizure detection in wearable EEG devices with gradient-boosted tree classifiers

The paper presents a combined seizure and artifact detection framework based on Gradient Boosted Trees. The framework achieves high accuracy in detecting seizures and artifacts, reducing false alarms. The algorithms are optimized for a Parallel Ultra-Low Power platform, enabling extended monitoring with a long battery lifespan. The paper highlights the benefits of integrating artifact detection in wearable epilepsy monitoring devices.

Thorir Mar Ingolfsson, Andrea Cossettini, Simone Benatti, Luca Benini

June 2022 In EMBC

Energy-Efficient Tree-Based EEG Artifact Detection

In this paper, we present implementations of energy-efficient artifact detection algorithms on a parallel ultra-low power platform.

Thorir Mar Ingolfsson, Mark Vero, Xiaying Wang, Lorenzo Lamberti, Luca Benini, Matteo Spallanzani

May 2022 In CF22

Reducing Neural Architecture Search Spaces with Training-Free Statistics and Computational Graph Clustering

In this paper, we present Clustering-Based REDuction (C-BRED), a new technique to reduce the size of NAS search spaces.

Thorir Mar Ingolfsson, Andrea Cossettini, Xiaying Wang, Enrico Tabanelli, Giuseppe Tagliavini, Philippe Ryvlin, Luca Benini, Simone Benatti

October 2021 In BIOCAS

Towards Long-term Non-invasive Monitoring for Epilepsy via Wearable EEG Devices

In this paper, we present implementations of seizure detection algorithms on a parallel ultra-low power platform.

Thorir Mar Ingolfsson, Xiaying Wang, Michael Hersche, Alessio Burrello, Lukas Cavigelli, Luca Benini

June 2021 In AICAS

ECG-TCN: Wearable Cardiac Arrhythmia Detection with a Temporal Convolutional Network

In this paper, we propose ECG-TCN, a novel temporal convolutional network (TCN) that achieves outstanding accuracy while requiring few trainable parameters.

Thorir Mar Ingolfsson, Michael Hersche, Xiaying Wang, Nobuaki Kobayashi, Lukas Cavigelli, Luca Benini

October 2020 In SMC

EEG-TCNet: An Accurate Temporal Convolutional Network for Embedded Motor-Imagery Brain–Machine Interfaces

In this paper, we propose EEG-TCNet, a novel temporal convolutional network (TCN) that achieves outstanding accuracy while requiring few trainable parameters.

Recent Publications

Quickly discover relevant content by filtering publications.

Contact

thoriri@iis.ee.ethz.ch
+41 78 251 26 44
Gloriastrasse 35, Zurich, 8092
Enter ETZ and take the stairs to Office 76.2 on Floor J
DM Me
Skype Me