Learning of lateral connections for representational invariant recognition

Authors:
Christian Keck;Jörg Lücke
Affiliations:
Frankfurt Institute for Advanced Studies, Goethe-Universität Frankfurt, Frankfurt am Main, Germany;Frankfurt Institute for Advanced Studies, Goethe-Universität Frankfurt, Frankfurt am Main, Germany
Venue:
ICANN'10 Proceedings of the 20th international conference on Artificial neural networks: Part III
Year:
2010

Citing 11
Cited 0

SEEMORE: combining color, shape, and texture histogramming in a neurally inspired approach to visual object recognition

Neural Computation
Rapid processing and unsupervised learning in a model of the cortical macrocolumn

Neural Computation
Maplets for correspondence-based object recognition

Neural Networks - 2004 Special issue: New developments in self-organizing systems
A fast learning algorithm for deep belief nets

Neural Computation
Maximal Causes for Non-linear Component Extraction

The Journal of Machine Learning Research
Rapid convergence to feature layer correspondences

Neural Computation
A Visual Object Recognition System Invariant to Scale and Rotation

ICANN '08 Proceedings of the 18th international conference on Artificial Neural Networks, Part I
Learning of Neural Information Routing for Correspondence Finding

ICANN '08 Proceedings of the 18th international conference on Artificial Neural Networks, Part II
A parallel computation that assigns canonical object-based frames of reference

IJCAI'81 Proceedings of the 7th international joint conference on Artificial intelligence - Volume 2
Receptive field self-organization in a model of the fine structure in v1 cortical columns

Neural Computation
Dynamics of cortical columns – self-organization of receptive fields

ICANN'05 Proceedings of the 15th international conference on Artificial Neural Networks: biological Inspirations - Volume Part I

Quantified Score

Hi-index	0.00

Visualization

Abstract

We study an artificial neural network that learns the invariance properties of objects from data. We start with a bag-of-features encoding of a specific object and repeatedly show the object in different transformations. The network then learns unsupervised from the data what the possible transformations are and what feature arrangements are typical for the object shown. The information about transformations and feature arrangements is hereby represented by a lateral network of excitatory connections among units that control the information exchange between an input and a down-stream neural layer. We build up on earlier work in this direction that kept a close relation to novel anatomical and physiological data on the cortical architecture and on its information processing and learning. At the same time we show, based on a new synaptic plasticity rules, that learning results in a strong increase of object finding rates in both artificial and more realistic experiments.