Babak Hosseini
I want to design effective machine learning algorithms which are interpretable for both humans and machines.
I am a Researcher in the Pattern Recognition Lab. of TU Dortmund University working with Gernot Fink. My research focus is design and implementation of advanced machine learning algorithms for specialized embedded devices.
I did my Ph.D. study in the Machine Learning Lab. of Cognitive Interaction Technology Center (CITEC) at Bielefeld University under the supervision of Barbara Hammer. The subject of my Ph.D. project was Semantic analysis of motion data.
My academic background is in Machine learning, Robotics, and Control Theory, and I have vocational experience in industrial sectors as a control/intelligent systems engineer. The topic of my Master dissertation was Concept Learning and Transfer among Heterogeneous Agents.
I was fortunate to work with Majid Nili and Babak N. Araabi at the University of Tehran, and with Ali K. Sedigh at the K. N. Toosi University.
News
Oct 2019: Just started my work in Pattern Recognition group of TU Dortmund University
Aug 2019: Deep-Aligned Convolutional Neural Network for Skeleton-based Action Recognition and Segmentation is accepted at ICDM 2019
Aug 2019: Interpretable Multiple-Kernel Prototype Learning for Discriminative Representation and Feature Selection is accepted at CIKM 2019
June 2019: Interpretable Discriminative Dimensionality Reduction and Feature Selection on the Manifold is accepted at ECML 2019
March 2019: Large-Margin Multiple Kernel Learning for Discriminative Features Selection and Representation Learning is accepted at IJCNN 2019
Multiple-Kernel Dictionary Learning for Reconstruction and Clustering of Unseen Multivariate Time-series is accepted at ESANN 2019.
Confident Kernel Sparse Coding and Dictionary Learning is accepted at ICDM 2018.
Non-Negative Local Sparse Coding for Subspace Clustering is accepted at IDA 2018.
Feasibility Based Large Margin Nearest Neighbor Metric Learning is accepted at ESANN 2018.
Talk on Non-negative Kernel Sparse Coding Frameworks for Efficient Analysis of Motion Data at BMVA Symposium on Human Activity Recognition and Monitoring 2017.
Non-Negative Kernel Sparse Coding for Analysis of Motion Data is accepted at ICANN 2016.
Efficient Metric Learning for the Analysis of Motion Data is accepted at DSAA 2015.
Research Interests
Interpretable Machine Learning
Deep learning
Time-series analysis
Applied Machine Learning
Kernel-learning
Education
PhD, Computer Science 2019 (expected)
Intelligent Systems PhD program
Bielefeld University
MSc, Control Engineering 2009
Focus on Machine Learning and Robotics
University of Tehran
BSc, Control Engineering 2006
K. N. Toosi University
Selected Publications and Projects
Kernel Based Dictionary Learning for Discriminative Representation of Multivariate Time-series
Babak Hosseini, Francois Petitjean, Germain Forestier, Barbara Hammer.
Working article
Feasibility Based Large Margin Nearest Neighbor Metric Learning
Babak Hosseini, Barbara Hammer.
ESANN 2018, Bruges.
[Paper]
Abstract Concept Learning Approach Based on Behavioural Feature Extraction
Babak Hosseini, Majid Nili, Babak N. Araabi
ICCEE 2009, Dubai.