Me Martin Längkvist
PhD in Information Technology
Msc Applied Physics and Electrical Engineering
Member of Applied Autonomous Sensor Systems
Machine Perception and Interaction Lab
School of Science and Technology
Örebro University, Sweden
Email:

Research Goal

My research interest is in machine learning, specifically learning good representations from raw sensory data.
I believe finding good representations is the key to designing a system that can solve interesting challenging
real-world problems, go beyond human-level intelligence, and ultimately explain complicated data for us that we
don't understand. In order to achieve this, I envision a learning algorithm that can learn feature representations
from both unlabeled and labeled data, be guided with and without human interaction, and that are on different
levels of abstractions in order to bridge the gap between low-level sensory data and high-level abstract concepts.

Google Scholar Page
Research Gate Page

Publications

Interactive Learning with Convolutional Neural Networks for Image Labeling [Published version], [bibtex]
Martin Längkvist, Marjan Alirezaie, Andrey Kiselev, and Amy Loutfi
IJCAI workshop on Interactive Machine Learning, 2016

Classification and Segmentation of Satellite Orthoimagery Using Convolutional Neural Networks [Published version], [code], [bibtex]
Martin Längkvist, Andrey Kiselev, Marjan Alirezaie, and Amy Loutfi
Remote Sens. 2016, 8(4), 329; doi:10.3390/rs8040329

Modeling Time-Series with Deep Networks [pdf]
Martin Längkvist
PhD thesis, Örebro University, 2014

Learning Feature Representations with a Cost-Relevant Sparse Autoencoder [Accepted version], [bibtex]
Martin Längkvist and Amy Loutfi
Int. J. Neur. Syst. 25, 1450034 (2015)

A Review of Unsupervised Feature Learning and Deep Learning for Time-Series Modeling [Published version], [Accepted version], [bibtex]
Martin Längkvist, Lars Karlsson, and Amy Loutfi
Pattern Recognition Letters, 2014, Volume 42, Pages 11-24

Fast Classification of Meat Spoilage Markers Using Nanostructured ZnO Thin Films and Unsupervised Feature Learning [Published version], [bibtex]
Martin Längkvist, Silvia Coradeschi, Amy Loutfi, and John Bosco Balaguru Rayappan
Sensors 2013, 13(2), 1578-1592; doi:10.3390/s130201578

Not all signals are created equal: Dynamic Objective Auto-Encoder for Multivariate Data [Accepted version], [poster], [data], [bibtex]
Martin Längkvist and Amy Loutfi
NIPS workshop on Deep Learning and Unsupervised Feature Learning, 2012

Sleep Stage Classification using Unsupervised Feature Learning [Published version], [data], [code], [bibtex]
Martin Längkvist, Lars Karlsson, and Amy Loutfi
Advances in Artificial Neural Systems, vol. 2012, Article ID 107046, 9 pages, 2012. doi:10.1155/2012/107046

Unsupervised feature learning for electronic nose data applied to Bacteria Identification in Blood [Accepted version], [poster], [data (blood)], [data (agar)], [bibtex]
Martin Längkvist and Amy Loutfi
NIPS workshop on Deep Learning and Unsupervised Feature Learning, 2011