History of pySPACE

pySPACE was founded in December 2012 as a unification and renewal of its two predecessors BOR (Benchmarking on Rails) and aBRI (adaptive Brain Reading Interface). These two frameworks originated during the VI-Bot project in 2008 at the German Research Center for Artificial Intelligence (DFKI GmbH), Robotics Innovation Center, Bremen (DFKI RIC). Still today these two foundations of pySPACE are clearly visible in its functionality.

Both, aBRI and BOR, were started by Jan Hendrik Metzen and Timo Duchrow. The task of BOR was to provide parallelization of independent processes to enable quick comparison of algorithms and parameters on large scales. For this, interfaces to other frameworks such as WEKA and MMLF were implemented (so called BOR operations) and BOR distributed the task when possible and intended by the user. The parallelization part, the pySPACE environment and interfacing using operations in pySPACE therefore has its roots in the BOR framework.

The task of aBRI was to provide quick and modular signal processing and it was purely used with EEG data (EEG=electroencephalogram). That’s where its name came from: the application. At the beginning aBRI was a wrapper around the MDP framework, but soon many own node implementations were implemented and large parts of the MDP functionality overwritten by own, better fitting functions. The main differences to MDP were that data in aBRI were two-dimensional and aBRI was processing each single data instance and did unlike MDP not include a whole dataset at once. aBRI had no own execution part, input and output of the data were defined elsewhere. That is why aBRI was inseparably connected with BOR from its very beginning. BOR took over the execution part of aBRI for offline analysis of the data. Another execution mode possible in aBRI was an online mode, where data could be directly streamed and processed. So, in pySPACE the signal processing capability, the node concept with its familiarity to MDP and the online mode (which is now pySPACE-live) comes from aBRI.

With the ending of the project VI-Bot and the beginning of the project IMMI at the DFKI RIC and the Robotics Group at the University of Bremen in May 2010, more and more nodes and operations were added, classification algorithms introduced to aBRI, and documentation and usability enhanced. Both frameworks were more and more connected and structure and style were constantly changing (even diverging in some parts). More functionality, like adaptivity, became a real part of the software in IMMI and the backend to the cluster software LoadLeveler was implemented.

The idea of unifying both frameworks under what is now pySPACE came into play, because we decided at some point to have a complete software with full execution support which still has the property of being modular and usable from outside its own environment. We took the opportunity to create an easy and transparent structure for both, users and developers, and now can make full use of all features of both frameworks and more. In the process of creating pySPACE, both structures were merged and completely rearranged. Furthermore, pySPACE is now fully independent of MDP which simplifies future developments and integration of new components.

We hope, you like our child.

Publications

pySPACE and its predecessors have been used in several publications as listed below. If you have your own publication using pySPACE please tell us, such that we can complete it.

main pySPACE paper:

Mario Michael Krell, Sirko Straube, Anett Seeland, Hendrik Wöhrle, Johannes Teiwes, Jan Hendrik Metzen, Elsa Andrea Kirchner, Frank Kirchner (2013)
In Frontiers in Neuroinformatics 7(40), doi: 10.3389/fninf.2013.00040

Journal Publications

PhD thesis with a lot of evaluations, implemented and published with pySPACE

Mario Michael Krell (2015)
PhD Thesis, University of Bremen, Bremen

introduction of new visualization approach for processing pipelines in pySPACE

Sirko Straube, Mario Michael Krell (2015)
Advances in Data Analysis and Classification, doi: 10.1007/s11634-015-0229-3

large comparison of online learning approaches and introduction of adaptive xDAWN

Hendrik Wöhrle, Mario Michael Krell, Sirko Straube, Su Kyoung Kim, Elsa Andrea Kirchner, Frank Kirchner (2015)
IEEE Transactions on Biomedical Engineering 62(7), doi: 10.1109/TBME.2015.2402252

evaluation of one class approaches including a variant of the BRMM classifier:

Mario Michael Krell, Hendrik Wöhrle (2015)
Pattern Recognition Letters, Elsevier, doi: 10.1016/j.patrec.2014.11.008

comparison of evaluation metrics for classification on data with imbalanced class ratio:

Sirko Straube, Mario Michael Krell (2014)
In Frontiers in Computational Neuroscience 8(43), doi:10.3389/fncom.2014.00043

evaluation of Balanced Relative Margin Machine (BRMM) classifier:

Mario Michael Krell, David Feess, Sirko Straube (2014)
In Pattern Recognition Letters 41, Elsevier, doi: 10.1016/j.patrec.2013.09.018

comparison of eye artifact removal methods:

Foad Ghaderi, Su Kyoung Kim, Elsa Andrea Kirchner (2014)
In Journal of Neuroscience Methods 221(0): 41-47, doi: 10.1016/j.jneumeth.2013.08.025

applications for brain-computer interfaces:

Elsa Andrea Kirchner, Su Kyoung Kim, Sirko Straube, Anett Seeland, Hendrik Wöhrle, Mario Michael Krell, Marc Tabie, Manfred Fahle (2013)
In PLoS ONE 8(12): e81732, doi:10.1371/journal.pone.0081732

large scale evaluation and comparison of sensor selection algorithms:

David Feess, Mario Michael Krell*, Jan Hendrik Metzen (2013)
In PLoS ONE 8(7): e67543, doi:10.1371/journal.pone.0067543

Conference Publications

evaluation of different online training dataset manipulations for online SVMs

Mario Michael Krell, Nils Wilshusen, Andrei Cristian Ignat, Su Kyoung Kim (2015)
Proceedings of the International Congress on Neurotechnology, Electronics and Informatics, SciTePress, doi: 10.5220/0005650700590067

evaluation of the regularized version of the adaptive xDAWN spatial filter

Mario Michael Krell, Anett Seeland, Hendrik Wöhrle
Proceedings of the International Congress on Neurotechnology, Electronics and Informatics, SciTePress, 2015

application in movement prediction with pySPACE live:

Anett Seeland, Hendrik Wöhrle, Sirko Straube, Elsa Andrea Kirchner (2013)
Online Movement Prediction in a Robotic Application Scenario
In 6th International IEEE EMBS Conference on Neural Engineering (NER): 41-44

evaluation of online classifiers:

Hendrik Wöhrle, Johannes Teiwes, Mario Michael Krell, Elsa Andrea Kirchner, Frank Kirchner (2013)
A Dataflow-Based Mobile Brain Reading System on Chip with Supervised Online Calibration
In International Congress on Neurotechnology, Electronics and Informatics, (NEUROTECHNIX-2013), SciTePress Digital Library

comparison of different classification score postprocessing methods for movement prediction:

Sirko Straube, Anett Seeland, David Feess (2013)
Striving for better and earlier movement prediction by postprocessing of classification scores
In International Congress on Neurotechnology, Electronics and Informatics, (NEUROTECHNIX-2013), SciTePress Digital Library

application in brain-computer interface for exoskeleton control:

Elsa Andrea Kirchner, Jan Albiez, Anett Seeland, Mathias Jordan, Frank Kirchner (2013)
Towards Assistive Robotics for Home Rehabilitation
In Proceedings of the 6th International Conference on Biomedical Electronics and Devices (BIODEVICES-13), SciTePress, 168-177

error potential detection for brain-computer interface:

Su Kyoung Kim, Elsa Andrea Kirchner (2013)
Classifier Transferability in the Detection of Error Related Potentials from Observation to Interaction
In Proceedings of the IEEE International Conference on Systems, Man, and Cybernetics

evaluation of adaptive periodic spatial filter (PiSF):

Foad Ghaderi, Sirko Straube (2013)
An adaptive and efficient spatial filter for event-related potentials
In Proceedings of European Signal Processing Conference (EUSIPCO)

evaluation of periodic spatio spectral filter:

Foad Ghaderi (2013)
In IEEE International Workshop on Machine Learning for Signal Processing (MLSP)

first paper about the periodic spatial filter (PiSF)

Foad Ghaderi, Elsa Andrea Kirchner, 2013
In Proceedings of the 10th IASTED International Conference on Biomedical Engineering (BioMed-2013)

classification in compressed space:

Yohannes Kassahun, Hendrik Wöhrle, Alexander Fabisch, Marc Tabie (2012)
In Artificial Neural Networks and Machine Learning – ICANN 2012, Lecture Notes in Computer Science: 108-115, doi: 978-3-642-33266-1_14

comparison of different downsampling methods and band pass filters for LRP:

Michele Folgheraiter, Elsa Andrea Kirchner, Anett Seeland, Su Kyoung Kim, Mathias Jordan, Hendrik Wöhrle, Bertold Bongardt, Steffen Schmidt, Jan Albiez, Frank Kirchner (2011)
A multimodal brain-arm interface for operation of complex robotic systems and upper limb motor recovery
In Proceedings of the 4th International Conference on Biomedical Electronics and Devices (BIODEVICES-11): 150-162

analysis of transferability of spatial filters:

Jan Hendrik Metzen, Su Kyoung Kim, Timo Duchrow, Elsa Andrea Kirchner, Frank Kirchner (2011)
In Proceedings of the 2011 IEEE Workshop on Statistical Signal Processing: 797-800, doi: 10.1109/SSP.2011.5967825

threshold adaptation:

Jan Hendrik Metzen, Elsa Andrea Kirchner (2011)
In Proceedings of the 2011 Conference of the German Classification Society (GfKl-2011): 138

ensemble classification for brain-computer interface:

Jan Hendrik Metzen, Su Kyoung Kim, Elsa Andrea Kirchner (2011)
In Pattern Recognition, Lecture Notes in Computer Science 6835: 366-375, doi: 10.1007/978-3-642-23123-0_37

application in Brain Reading:

Elsa Andrea Kirchner, Hendrik Wöhrle, Constantin Bergatt, Su Kyoung Kim, Jan Hendrik Metzen, David Feess, Frank Kirchner (2010)
In Proceedings of the 10th International Symposium on Artificial Intelligence, Robotics and Automation in Space: 448-455

Other Publications

pySPACE has been presented at the NIPS2013 workshop Machine Learning Open Source Software: Towards Open Workflows.