Papers by Benjamin Auffarth
This thesis deals with machine learning techniques for the extraction of structure and the analys... more This thesis deals with machine learning techniques for the extraction of structure and the analysis of the vertebrate olfactory pathway based on related methods. Some of its main contributions are summarized below. We have performed a systematic investigation for classification in biomedical images with the goal of recognizing a material in these images by its texture. This investigation included (i) different measures for evaluating the importance of image descriptors (features), (ii) methods to select a feature set based on these evaluations, and (iii) classification algorithms. Image features were evaluated according to their estimated relevance for the classification task and their redundancy with other features. For this purpose, we proposed a framework for relevance and redundancy measures and, within this framework, we proposed two new measures. These were the value difference metric and the fit criterion. Both measures performed well in comparison with other previously used ...
by a genetic algorithm with biased mutation operator
POSTER PRESENTATION Open Access
KSHV encoded miRNA single nucleotide polymorphisms identified in clinical samples can affect miRN... more KSHV encoded miRNA single nucleotide polymorphisms identified in clinical samples can affect miRNA processing and level of expression
doi: 10.3389/fnsys.2011.00084 Map formation in the olfactory bulb by axon guidance of olfactory n... more doi: 10.3389/fnsys.2011.00084 Map formation in the olfactory bulb by axon guidance of olfactory neurons
Frontiers in Systems Neuroscience, 2011
The relationship between molecular properties of odorants and neural activities is arguably one o... more The relationship between molecular properties of odorants and neural activities is arguably one of the most important issues in olfaction and the rules governing this relationship are still not clear. In the olfactory bulb (OB), glomeruli relay olfactory information to second-order neurons which in turn project to cortical areas. We investigate relevance of odorant properties, spatial localization of glomerular coding sites, and size of coding zones in a dataset of [ 14 C] 2-deoxyglucose images of glomeruli over the entire OB of the rat. We relate molecular properties to activation of glomeruli in the OB using a non-parametric statistical test and a support-vector machine classification study. Our method permits to systematically map the topographic representation of various classes of odorants in the OB. Our results suggest many localized coding sites for particular molecular properties and some molecular properties that could form the basis for a spatial map of olfactory information. We found that alkynes, alkanes, alkenes, and amines affect activation maps very strongly as compared to other properties and that amines, sulfur-containing compounds, and alkynes have small zones and high relevance to activation changes, while aromatics, alkanes, and carboxylics acid recruit very big zones in the dataset. Results suggest a local spatial encoding for molecular properties.
UNIVERSITAET OSNABRUECK, 2006
Hiermit erkläre ich, Benjamin Joachim Auffarth, die vorliegende Arbeit Investigation into differe... more Hiermit erkläre ich, Benjamin Joachim Auffarth, die vorliegende Arbeit Investigation into differences of monkey and human eye movements selbständig verfasst zu haben und keine anderen Quellen oder Hilfsmittel als die angegebenen verwendet zu haben. Osnabrück, den 25.11.2006
The olfactory bulb (OB) receives input from the Olfactory Receptor Neurons through a neuropil lay... more The olfactory bulb (OB) receives input from the Olfactory Receptor Neurons through a neuropil layer composed of glomeruli. Each glomeruli is able to detect certain molecular properties of the odorant and similar properties seem to be detected by nearby glomeruli. As a consequence, the presence of certain molecular properties shows in the activation of zones in the OB which are sometimes called "modules" (Hildebrand and Shepherd 1997, Mori et al. 1999, Leon and Johnson 2003). We analyzed experimental image data of the rat olfactory bulb in order to determine where representational sites of properties are located. Using c-means clustering we partitioned the bulb into different zones, from which we wanted to determine those which could qualify as modules.
With the goal of deepen in the understanding of coding of chemical information in the olfactory s... more With the goal of deepen in the understanding of coding of chemical information in the olfactory system, a large data set consisting of rat's olfactory bulb activity values in response to several different volatile compounds has been analyzed by fuzzy c-means clustering methods. Clustering should help to discover groups of glomeruli that are similary activated according to their response profiles across the odorants. To investigate the significance of the achieved fuzzy partitions we developed and applied a novel validity approach based on cluster stability. Our results show certain level of glomerular clustering in the olfactory bulb and indicate that exist a main chemo-topic subdivision of the glomerular layer in few macro-area which are rather specific to particular functional groups of the volatile molecules.
Frontiers in Systems Neuroscience, 2011
doi: 10.3389/fnsys.2011.00082 Continuous spatial representations in the olfactory bulb may reflec... more doi: 10.3389/fnsys.2011.00082 Continuous spatial representations in the olfactory bulb may reflect perceptual categories
doi: 10.3389/fnsys.2011.00082 Continuous spatial representations in the olfactory bulb may reflec... more doi: 10.3389/fnsys.2011.00082 Continuous spatial representations in the olfactory bulb may reflect perceptual categories
This thesis deals with machine learning techniques for the extraction of structure and the analys... more This thesis deals with machine learning techniques for the extraction of structure and the analysis of the vertebrate olfactory pathway based on related methods. Some of its main contributions are ...
2005 IEEE International Conference on Intelligent Engineering Systems, 2005. INES '05., 2005
Image processing of natural scenes is very processingintensive. Through the localization of salie... more Image processing of natural scenes is very processingintensive. Through the localization of salient regions, later recognition processes can take place more efficiently, by focusing computation-intensive processing on several areas. Our approach to designing an artificial visual system is inspired by early filtering mechanisms in the human visual system. Our technique of calculating salient regions is computationally efficient and flexible, and can be extended to other applications.
PLOS ONE, 2015
Odours are highly complex, relying on hundreds of receptors, and people are known to disagree in ... more Odours are highly complex, relying on hundreds of receptors, and people are known to disagree in their linguistic descriptions of smells. It is partly due to these facts that, it is very hard to map the domain of odour molecules or their structure to that of perceptual representations, a problem that has been referred to as the Structure-Odour-Relationship. We collected a number of diverse open domain databases of odour molecules having unorganised perceptual descriptors, and developed a graphical method to find the similarity between perceptual descriptors; which is intuitive and can be used to identify perceptual classes. We then separately projected the physico-chemical and perceptual features of these molecules in a non-linear dimension and clustered the similar molecules. We found a significant overlap between the spatial positioning of the clustered molecules in the physico-chemical and perceptual spaces. We also developed a statistical method of predicting the perceptual qualities of a novel molecule using its physico-chemical properties with high receiver operating characteristics(ROC).
Case studies in neural networks and an exploration into data reduction
We describe the application of learining with neural networks and support vector machines on diff... more We describe the application of learining with neural networks and support vector machines on different data sets. We report high results for hepatitis, sonar, and ionosphere, fail completely with a eye tracking data set, and for a natural language learning data set, after desastrous clas sification rates with more conventional kernel methods, MLPs, and others, we obtain good results with custom kernels that take into account the discrete nature of features. We conclude that the use of learning methods differed very much over data sets. At last, we present a method of unsupervised feature selection based on the unmixing matrix of ICA, which we found promising.
Watson & Trotzky, 2006) describes the implementation of a distributed case-based reasoning system... more Watson & Trotzky, 2006) describes the implementation of a distributed case-based reasoning system, called case-agent system to mean computational agents representing cases, contrasting it to the conventional approach of one process (agent) that accesses a single case-base. The authors aim to show feature modelling in their system and test performance of retrieval as compared to other retrieval methods, such as based on k-nearest neighbor, compressed nearest neighbor, and the footprint-based retrieval (Smyth & McKenna, 1999). They also demonstrate a simple visualization and conclude that their system's performance characteristics is similar to footprintbased retrieval, and-due to distributedness and emergent properties-has promising applications. In a discussion, I criticize many aspects of the paper and conclude that the article has some serious shortcomings.
We implemented two auction frameworks in 3-APL in order to test its capabilities. We provide some... more We implemented two auction frameworks in 3-APL in order to test its capabilities. We provide some criticism, including concerning the IDE, the parser, the lack of more sophisticated data structures. We conclude that experience has been quite good, and we think that, as it is noted before, 3-APL could be a great language for BDI agents if this kind of problems were solved.
Uploads
Papers by Benjamin Auffarth