Refine
Year of publication
- 2014 (22) (remove)
Document Type
- Doctoral Thesis (7)
- Article (6)
- Preprint (5)
- Working Paper (2)
- Book (1)
- Part of a Book (1)
Language
- English (22) (remove)
Has Fulltext
- yes (22)
Is part of the Bibliography
- no (22)
Keywords
- Big Data (1)
- Computational Humanities (1)
- Computer Science (1)
- Digital Humanities (1)
- Energy Efficiency (1)
- Experimental Methods (1)
- Frontmatter (1)
- GPGPU (1)
- GPU algorithms (1)
- Green Computing (1)
Institute
- Informatik (22) (remove)
Efficient algorithms for object recognition are crucial for the newly robotics and computer vision applications that demand real-time and on-line methods. Some examples are autonomous systems, navigating robots, autonomous driving. In this work, we focus on efficient semantic segmentation, which is the problem of labeling each pixel of an image with a semantic class.
Our aim is to speed-up all of the parts of the semantic segmentation pipeline. We also aim at delivering a labeling solution on a time budget, that can be decided on-the-fly. For this purpose, we analyze all the components of the semantic segmentation pipeline, and identify the computational bottleneck of each of them. The different components of the pipeline are over-segmenting the image with local regions, extracting features and classify the local regions, and the final inference of the image labeling with semantic classes. We focus on each of these steps.
First, we introduce a new superpixel algorithm to over-segment the image. Our superpixel method runs in real-time and can deliver a solution at any time budget. Then, for feature extraction, we focus on the framework that computes descriptors and encodes them, followed by a pooling step. We see that the encoding step is the bottleneck, for computational efficiency and performance. We present a novel assignment-based encoding formulation, that allows for the design of a new, very efficient, encoding. Finally, the image labeling output is obtained modeling the dependencies with a Conditional Random Field (CRF). In semantic image segmentation, the computational cost of instantiating the potentials is much higher than MAP inference. We introduce Active MAP inference to on-the-fly select a subset of potentials to be instantiated in the energy function, leaving the rest as unknown, and to estimate the MAP labeling from such incomplete energy function.
We perform experiments on all proposed methods for the different parts of the semantic segmentation pipeline. We show that our superpixel extraction achieves higher accuracy than state-of-the-art on standard superpixel benchmark, while it runs in real-time. We test our feature encoding on standard image classification and segmentation benchmarks, and we show that our method achieves competitive results with the state-of-the-art, and requires less time and memory. Finally, results for semantic segmentation benchmark show that Active MAP inference achieves similar levels of accuracy but with major efficiency gains.
Acceleration of Biomedical Image Processing and Reconstruction with FPGAs
Increasing chip sizes and better programming tools have made it possible to increase the boundaries of application acceleration with reconfigurable computer chips. In this thesis the potential of acceleration with Field Programmable Gate Arrays (FPGAs) is examined for applications that perform biomedical image processing and reconstruction. The dataflow paradigm was used to port the analysis of image data for localization microscopy and for 3D electron tomography from an imperative description towards the FPGA for the first time.
After the primitives of image processing on FPGAs are presented, a general workflow is given for analyzing imperative source code and converting it to a hardware pipeline where every node processes image data in parallel. The theoretical foundation is then used to accelerate both example applications. For localization microscopy, an acceleration of 185 compared to an Intel i5 450 CPU was achieved, and electron tomography could be sped up by a factor of 5 over an Nvidia Tesla C1060 graphics card while maintaining full accuracy in both cases.
Research in the field of Digital Humanities, also known as Humanities Computing, has seen a steady increase over the past years. Situated at the intersection of computing science and the humanities, present efforts focus on making resources such as texts, images, musical pieces and other semiotic artifacts digitally available, searchable and analysable. To this end, computational tools enabling textual search, visual analytics, data mining, statistics and natural language processing are harnessed to support the humanities researcher. The processing of large data sets with appropriate software opens up novel and fruitful approaches to questions in the traditional humanities. This report summarizes the Dagstuhl seminar 14301 on “Computational Humanities - bridging the gap between Computer Science and Digital Humanities”.
1998 ACM Subject Classification I.2.7 Natural Language Processing, J.5 Arts and Humanities
Data structures and advanced models of computation on big data : report from Dagstuhl seminar 14091
(2014)
This report documents the program and the outcomes of Dagstuhl Seminar 14091 "Data Structures and Advanced Models of Computation on Big Data". In today's computing environment vast amounts of data are processed, exchanged and analyzed. The manner in which information is stored profoundly influences the efficiency of these operations over the data. In spite of the maturity of the field many data structuring problems are still open, while new ones arise due to technological advances.
The seminar covered both recent advances in the "classical" data structuring topics as well as new models of computation adapted to modern architectures, scientific studies that reveal the need for such models, applications where large data sets play a central role, modern computing platforms for very large data, and new data structures for large data in modern architectures.
The extended abstracts included in this report contain both recent state of the art advances and lay the foundation for new directions within data structures research.
1D-3D hybrid modeling : from multi-compartment models to full resolution models in space and time
(2014)
Investigation of cellular and network dynamics in the brain by means of modeling and simulation has evolved into a highly interdisciplinary field, that uses sophisticated modeling and simulation approaches to understand distinct areas of brain function. Depending on the underlying complexity, these models vary in their level of detail, in order to cope with the attached computational cost. Hence for large network simulations, single neurons are typically reduced to time-dependent signal processors, dismissing the spatial aspect of each cell. For single cell or networks with relatively small numbers of neurons, general purpose simulators allow for space and time-dependent simulations of electrical signal processing, based on the cable equation theory. An emerging field in Computational Neuroscience encompasses a new level of detail by incorporating the full three-dimensional morphology of cells and organelles into three-dimensional, space and time-dependent, simulations. While every approach has its advantages and limitations, such as computational cost, integrated and methods-spanning simulation approaches, depending on the network size could establish new ways to investigate the brain. In this paper we present a hybrid simulation approach, that makes use of reduced 1D-models using e.g., the NEURON simulator—which couples to fully resolved models for simulating cellular and sub-cellular dynamics, including the detailed three-dimensional morphology of neurons and organelles. In order to couple 1D- and 3D-simulations, we present a geometry-, membrane potential- and intracellular concentration mapping framework, with which graph- based morphologies, e.g., in the swc- or hoc-format, are mapped to full surface and volume representations of the neuron and computational data from 1D-simulations can be used as boundary conditions for full 3D simulations and vice versa. Thus, established models and data, based on general purpose 1D-simulators, can be directly coupled to the emerging field of fully resolved, highly detailed 3D-modeling approaches. We present the developed general framework for 1D/3D hybrid modeling and apply it to investigate electrically active neurons and their intracellular spatio-temporal calcium dynamics.
Mathematical modeling of Arabidopsis thaliana with focus on network decomposition and reduction
(2014)
Systems biology has become an important research field during the last decade. It focusses on the understanding of the systems which emit the measured data. An important part of this research field is the network analysis, investigating biological networks. An essential point of the inspection of these network models is their validation, i.e., the successful comparison of predicted properties to measured data. Here especially Petri nets have shown their usefulness as modeling technique, coming with sound analysis methods and an intuitive representation of biological network data.
A very important tool for network validation is the analysis of the Transition-invariants (TI), which represent possible steady-state pathways, and the investigation of the liveness property. The computational complexity of the determination of both, TI and liveness property, often hamper their investigation.
To investigate this issue, a metabolic network model is created. It describes the core metabolism of Arabidopsis thaliana, and it is solely based on data from the literature. The model is too complex to determine the TI and the liveness property.
Several strategies are followed to enable an analysis and validation of the network. A network decomposition is utilized in two different ways: manually, motivated by idea to preserve the integrity of biological pathways, and automatically, motivated by the idea to minimize the number of crossing edges. As a decomposition may not be preserving important properties like the coveredness, a network reduction approach is suggested, which is mathematically proven to conserve these important properties. To deal with the large amount of data coming from the TI analysis, new organizational structures are proposed. The liveness property is investigated by reducing the complexity of the calculation method and adapting it to biological networks.
The results obtained by these approaches suggest a valid network model. In conclusion, the proposed approaches and strategies can be used in combination to allow the validation and analysis of highly complex biological networks.
The morphology of presynaptic specializations can vary greatly ranging from classical single-release-site boutons in the central nervous system to boutons of various sizes harboring multiple vesicle release sites. Multi-release-site boutons can be found in several neural contexts, for example at the neuromuscular junction (NMJ) of body wall muscles of Drosophila larvae. These NMJs are built by two motor neurons forming two types of glutamatergic multi-release-site boutons with two typical diameters. However, it is unknown why these distinct nerve terminal configurations are used on the same postsynaptic muscle fiber. To systematically dissect the biophysical properties of these boutons we developed a full three-dimensional model of such boutons, their release sites and transmitter-harboring vesicles and analyzed the local vesicle dynamics of various configurations during stimulation. Here we show that the rate of transmission of a bouton is primarily limited by diffusion-based vesicle movements and that the probability of vesicle release and the size of a bouton affect bouton-performance in distinct temporal domains allowing for an optimal transmission of the neural signals at different time scales. A comparison of our in silico simulations with in vivo recordings of the natural motor pattern of both neurons revealed that the bouton properties resemble a well-tuned cooperation of the parameters release probability and bouton size, enabling a reliable transmission of the prevailing firing-pattern at diffusion-limited boutons. Our findings indicate that the prevailing firing-pattern of a neuron may determine the physiological and morphological parameters required for its synaptic terminals.
This volume contains the papers presented at the First International Workshop on Rewriting Techniques for Program Transformations and Evaluation (WPTE 2014) which was held on July 13, 2014 in Vienna, Austria during the Vienna Summer of Logic 2014 (VSL 2014) as a workshop of the Sixth Federated Logic Conference (FLoC 2014). WPTE 2014 was affiliated with the 25th International Conference on Rewriting Techniques and Applications joined with the 12th International Conference on Typed Lambda Calculi and Applications (RTA/TLCA 2014).
Local protein synthesis has re-defined our ideas on the basic cellular mechanisms that underlie synaptic plasticity and memory formation. The population of messenger RNAs that are localised to dendrites, however, remains sparsely identified. Furthermore, neuronal morphological complexity and spatial compartmentalisation require efficient mechanisms for messenger RNA localisation and control over translational efficiency or transcript stability. 3’ untranslated regions, downstream from stop codons, are recognised for providing binding platforms for many regulatory units, thus encoding the processing of the above processes. The hippocampus, a part of the brain involved in the formation, organisation and storage of memories, provides a natural platform to investigate patterns of RNA localisation. The hippocampus comprises tissue layers, which naturally separate the principle neuronal cell bodies from their processes (axons and dendrites). Identifying the full-complement of localised transcripts and associated 3’UTR isoforms is of great importance to understand both basic neuronal functions and principles of synaptic plasticity. These findings can be used to study the properties of neuronal networks as well as to understand how these networks malfunction in neuronal diseases.
Here, deep sequencing is used to identify the mRNAs resident in the synaptic neuropil in the hippocampus. Analysis of a neuropil data set yields a list of 8,379 transcripts of which 2,550 are localised in dendrites and/or axons. Using a fluorescent barcode strategy to label individual mRNAs shows that the relative abundance of different mRNAs in the neuropil varies over 5 orders of magnitude. High-resolution in situ hybridisation validated the presence of mRNAs in both cultured neurons and hippocampal slices. Among the many mRNAs identified, a large fraction of known synaptic proteins including signaling molecules, scaffolds and receptors is discovered. These results reveal a previously unappreciated enormous potential for the local protein synthesis machinery to supply, maintain and modify the dendritic and synaptic proteome.
Using advances in library preparation for next generation sequencing experiments, the diversity of 3’UTR isoforms present in localised transcripts from the rat hippocampus is examined. The obtained results indicate that there is an increase in 3’UTR heterogeneity and 3’UTR length in neuronal tissue. The evolutionary importance of the 3’UTR diversity and correlation with changes in species,tissue and cell complexity is investigated. The conducted analysis reveals the population of 3’UTR isoforms required for transcript localisation in overall neuronal transcriptome as well as the regulatory elements and binding sites specific for neuronal compartments. The configuration of poly(A) signals is correlated with gene function and can be further exploit to determine similar mechanisms for alternative polyadenylation.
Usage of custom specified methods for next-generation sequencing as well as novel approaches for RNA quantification and visualisation necessitate the development and implementation of new downstream analytic methods. Library methods for data-mining transcripts annotation, expression and ontology relations is provided. Usage of a specialised search engine targeting key features of previous experiments is proposed. A processing pipeline for NanoString technology, defining experimental quality and exploiting methods for data normalisation is developed. High-resolution in situ images are analysed by custom application, showing a correlation between RNA quantity and spatial distribution. The vast variety of bioinformatic methods included in this work indicates the importance of downstream analysis to reach biological conclusions. Maintaining the integrability and modularity of our implementations is of great priority, as the dynamic nature of many experimental techniques requires constant improvement in computational analysis.