TU Berlin

Neural Information ProcessingLearning on Structured Representations

Neuronale Informationsverarbeitung

Page Content

to Navigation

Learning on Structured Representations


Learning from examples in order to predict is one of the standard tasks in machine learning. Many techniques have been developed to solve classification and regression problems, but by far, most of them were specifically designed for vectorial data. Vectorial data are very convenient because of the structure imposed by the Euclidean metric. For many data sets (protein sequences, text, images, videos, chemical formulas, etc.) a vector-based description is not only inconvenient but may simply wrong, and representations that consider relationships between objects or that embed objects in spaces with non-Euclidean structure are often more appropriate. Here we follow different approaches to extend learning from examples to non-vectorial data. One approach is focussed on an extension of kernel methods leading to learning algorithms specifically designed for relational data representations of a general form. In a second approach - specifically designed for objects which are naturally represented in terms of finite combinatorial structures - we explore embeddings into quotient spaces of a Euclidean vector space ("structure spaces"). In a third approach we consider representations of in spaces with data adapted geometries, i.e. using Riemannian manifolds as models for data spaces. In this context we are also interested in active learning schemes which are based on geometrical concepts. The developed algorithms have been applied to various applications domains, including bio- and chemoinformatics (cf. "Research" page "Applications to Problems in Bio- and Chemoinformatics") and the analysis of multimodal neural data (cf. "Research" page "MRI, EM, Autoradiography, and Multi-modal Data").

Acknowledgement: This work was funded by the BMWA and by the Technical University of Berlin.


The Potential Support Vector Machine (P-SVM)

Selected Publications:

Extending Integrate-and-fire Model Neurons to Account for the Effects of Weak Electric Fields and Input Filtering Mediated by the Dendrite
Citation key Aspart2016
Author Aspart, F. and Ladenbauer, J. and Obermayer, K.
Pages e1005206
Year 2016
DOI DOI 10.1371/ journal.pcbi.1005206
Journal PLOS Computional Biology
Volume 12
Number 11
Abstract The collective dynamics of neuronal populations can be efficiently studied using single-compartment (point) model neurons of the integrate-and-fire (IF) type. Existing point neuron models are intrinsically not able to appropriately reproduce (i) the effects of dendrites on synaptic input integration or (ii) the modulation of neuronal activity due to an electric field, which strongly depends on the dendritic morphology. Weak electric fields, as generated endogenously or through transcranial electrical stimulation, have recently gained increased attention because of their ability to modulate ongoing neuronal activity. However, the underlying mechanisms are not well understood. Here, we extend the popular spiking point neuron model class to accurately reflect input filtering and weak electric field effects as present in a canonical spatially extended “ball-and-stick” (BS) neuron model. We analytically derive additional components for two major types of IF point neuron models to exactly reproduce the subthreshold somatic voltage dynamics of the BS model with arbitrary morphology exposed to an oscillating electric field. Also the spiking dynamics for suprathreshold fluctuating inputs is well reproduced by the extended point models. Through this approach we further show that the presence of a dendritic cable (i) attenuates the somatic subthreshold response to slowly-varying inputs and (ii) mediates spike rate resonance, or equivalently, pronounced spike to field coherence, in the beta and gamma frequency range due to an oscillatory weak electric field. Our point neuron model extension is simple to implement and well suited for studying the dynamics of populations with heterogeneous neuronal morphology and the effects of weak electric fields on population activity.
Bibtex Type of Publication Selected:main selected:structured selected:publications
Link to original publication Download Bibtex entry


Quick Access

Schnellnavigation zur Seite über Nummerneingabe