Time filter

Source Type

Baehrens D.,TU Berlin | Schroeter T.,TU Berlin | Harmeling S.,MPI for Biological Cybernetics | Kawanabe M.,Fraunhofer Institute Firstida | And 2 more authors.
Journal of Machine Learning Research | Year: 2010

After building a classifier with modern tools of machine learning we typically have a black box at hand that is able to predict well for unseen data. Thus, we get an answer to the question what is the most likely label of a given unseen data point. However, most methods will provide no answer why the model predicted a particular label for a single instance and what features were most influential for that particular instance. The only method that is currently able to provide such explanations are decision trees. This paper proposes a procedure which (based on a set of assumptions) allows to explain the decisions of any classification method. © 2010 David Baehrens, Timon Schroeter, Stefan Harmeling, Motoaki Kawanabe, Katja Hansen and Klaus-Robert Müller.

Gretton A.,MPI for Biological Cybernetics | Gretton A.,Carnegie Mellon University | Gyorfi L.,Budapest University of Technology and Economics
Journal of Machine Learning Research | Year: 2010

Three simple and explicit procedures for testing the independence of two multi-dimensional random variables are described. Two of the associated test statistics (L1, log-likelihood) are defined when the empirical distribution of the variables is restricted to finite partitions. A third test statistic is defined as a kernel-based independence measure. Two kinds of tests are provided. Distributionfree strong consistent tests are derived on the basis of large deviation bounds on the test statistics: these tests make almost surely no Type 1 or Type II error after a random sample size. Asymptotically α-level tests are obtained from the limiting distribution of the test statistics. For the latter tests, the Type I error converges to a fixed non-zero value a, and the Type II error drops to zero, for increasing sample size. All tests reject the null hypothesis of independence if the test statistics become large. The performance of the tests is evaluated experimentally on benchmark data. © 2010 Arthur Gretton and László Györfi.

Palm G.,University of Ulm | Knoblauch A.,Honda Corporation | Knoblauch A.,Albstadt-Sigmaringen University of Applied Sciences | Hauser F.,University of Ulm | Schuz A.,MPI for Biological Cybernetics
Biological Cybernetics | Year: 2014

Donald Hebb’s concept of cell assemblies is a physiology-based idea for a distributed neural representation of behaviorally relevant objects, concepts, or constellations. In the late 70s Valentino Braitenberg started the endeavor to spell out the hypothesis that the cerebral cortex is the structure where cell assemblies are formed, maintained and used, in terms of neuroanatomy (which was his main concern) and also neurophysiology. This endeavor has been carried on over the last 30 years corroborating most of his findings and interpretations. This paper summarizes the present state of cell assembly theory, realized in a network of associative memories, and of the anatomical evidence for its location in the cerebral cortex. © 2014, Springer-Verlag Berlin Heidelberg.

Celicanin Z.,University of Basel | Bieri O.,University of Basel | Preiswerk F.,University of Basel | Cattin P.,University of Basel | And 3 more authors.
Magnetic Resonance in Medicine | Year: 2015

Purpose: Respiratory organ motion is still the major challenge of various image-guided treatments in the abdomen. Dynamic organ motion tracking, necessary for the treatment control, can be performed with volumetric time-resolved MRI that sequentially acquires one image and one navigator slice. Here, a novel imaging method is proposed for truly simultaneous high temporal resolution acquisition. Methods: A standard balanced steady state free precession sequence was modified to simultaneously acquire two superimposed slices with different phase cycles, namely an image and a navigator slice. Instead of multiband RF pulses, two separate RF pulses were used for the excitation. Images were reconstructed using offline CAIPIRINHA reconstruction. Phantom and in vivo measurements of healthy volunteers were performed and evaluated. Results: Phantom and in vivo measurements showed good image quality with high signal-to-noise ratio (SNR) and no reconstruction issues. Conclusion: We present a novel imaging method for truly simultaneous acquisition of image and navigator slices for four-dimensional (4D) MRI of organ motion. In this method, the time lag between the sequential acquisitions is eliminated, leading to an improved accuracy of organ motion models, while CAIPIRINHA reconstruction results in an improved SNR compared with an existing 4D MRI approach. © 2014 Wiley Periodicals, Inc.

Celicanin Z.,University of Basel | Auboiroux V.,University of Geneva | Bieri O.,University of Basel | Petrusca L.,University of Geneva | And 5 more authors.
Magnetic Resonance in Medicine | Year: 2014

Purpose: Magnetic resonance-guided high-intensity focused ultrasound is considered to be a promising treatment for localized cancer in abdominal organs such as liver, pancreas, or kidney. Abdominal motion, anatomical arrangement, and required sustained sonication are the main challenges. Methods: MR acquisition consisted of thermometry performed with segmented gradient-recalled echo echo-planar imaging, and a segment-based one-dimensional MR navigator parallel to the main axis of motion to track the organ motion. This tracking information was used in real-time for: (i) prospective motion correction of MR thermometry and (ii) HIFU focal point position lock-on target. Ex vivo experiments were performed on a sheep liver and a turkey pectoral muscle using a motion demonstrator, while in vivo experiments were conducted on two sheep liver. Results: Prospective motion correction of MR thermometry yielded good signal-to-noise ratio (range, 25 to 35) and low geometric distortion due to the use of segmented EPI. HIFU focal point lock-on target yielded isotropic in-plane thermal build-up. The feasibility of in vivo intercostal liver treatment was demonstrated in sheep. Conclusion: The presented method demonstrated in moving phantoms and breathing sheep accurate motioncompensated MR thermometry and precise HIFU focal point lock-on target using only real-time pencil-beam navigator tracking information, making it applicable without any pretreatment data acquisition or organ motion modeling. © 2013 Wiley Periodicals, Inc.

Discover hidden collaborations