Justin Bayer

Most of the time, I do machine learning. Currently, I am pursuing a doctorate at the Technische Unversitaet Muenchen as a member of the chair for robotics and embedded systems, part of the department for informatics. I am part of the group for biomimetic robots and machine learning.

You can find me on the web giving answers at StackOverflow and CrossValidated. My open source activity can be seen at Github.

From time to time, I freelance. My specialities are sequential and time series data, visualizations and deep learning. I have worked in brain computer interfaces, biological signals, energy consumption and more areas.

The best way to reach me is via mail: bayer.justin@googlemail.com.

Research

Some articles came out of the scientific work I do. They are listed below.

• Training Neural Networks with Implicit Variance (Bayer, Justin; Osendorfer, Christian; Urban, Sebastian; van der Smagt, Patrick) International Conference on Neural Information Processing Systems, 2013.
• Convolutional Neural Networks learn compact local image descriptors (Osendorfer, C., Bayer, J., & van der Smagt, P.) arXiv preprint arXiv:1304.7948
• Unsupervised Feature Learning for local image descriptors (Osendorfer, C., Bayer, J., & van der Smagt, P.) arXiv preprint arXiv:1301.2840
• Learning sequence neighbourhood metrics (Bayer, J., Osendorfer, C., & van der Smagt, P. ) Artificial Neural Networks and Machine Learning–ICANN 2012 (pp. 531-538). Springer Berlin Heidelberg.
• Identification of human limb stiffness in 5 DoF and estimation via EMG. (Lakatos, D., Rüschen, D., Bayer, J., Vogel, J., & van der Smagt, P.) 13th International Symposium on Experimental Robotics June 17-21, 2012
• PyBrain (Schaul, T., Bayer, J., Wierstra, D., Sun, Y., Felder, M., Sehnke, F., ... & Schmidhuber, J.) The Journal of Machine Learning Research, 11, 743-746.
• Evolving memory cell structures for sequence learning (Bayer, J., Wierstra, D., Togelius, J., & Schmidhuber, J. ) In Artificial Neural Networks–ICANN 2009 (pp. 755-764). Springer Berlin Heidelberg.

Talks

• Python and 0MQ do scale to real time really fast robot control at the Munich data geek meetup, august edition.