TU Berlin

Neural Information ProcessingLearning on Structured Representations

Neuronale Informationsverarbeitung

Page Content

to Navigation

Learning on Structured Representations

Lupe

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.

Software:

The Potential Support Vector Machine (P-SVM)

Selected Publications:

pypet: A Python Toolkit for Data Management of Parameter Explorations
Citation key Meyer2016b
Author Meyer, R. and Obermayer, K.
Year 2016
DOI 10.3389/fninf.2016.00038
Journal Frontiers Neuroinformatics
Volume 10
Number 38
Abstract pypet (Python parameter exploration toolkit) is a new multi-platform Python toolkit for managing numerical simulations. Sampling the space of model parameters is a key aspect of simulations and numerical experiments. pypet is designed to allow easy and arbitrary sampling of trajectories through a parameter space beyond simple grid searches. pypet collects and stores both simulation parameters and results in a single HDF5 file. This collective storage allows fast and convenient loading of data for further analyses. pypet provides various additional features such as multiprocessing and parallelization of simulations, dynamic loading of data, integration of git version control, and supervision of experiments via the electronic lab notebook Sumatra. pypet supports a rich set of data formats, including native Python types, Numpy and Scipy data, Pandas DataFrames, and BRIAN(2) quantities. Besides these formats, users can easily extend the toolkit to allow customized data types. pypet is a flexible tool suited for both short Python scripts and large scale projects. pypet's various features, especially the tight link between parameters and results, promote reproducible research in computational neuroscience and simulation-based disciplines.
Bibtex Type of Publication Selected:main selected:structured
Link to original publication Download Bibtex entry

Navigation

Quick Access

Schnellnavigation zur Seite über Nummerneingabe