8 research outputs found
Unsupervised learning in probabilistic neural networks with multi-state metal-oxide memristive synapses
No full textIn an increasingly data-rich world the need for developing computing systems that cannot only process, but ideally also interpret big data is becoming continuously more pressing. Brain-inspired concepts have shown great promise towards addressing this need. Here we demonstrate unsupervised learning in a probabilistic neural network that utilizes metal-oxide memristive devices as multi-state synapses. Our approach can be exploited for processing unlabelled data and can adapt to time-varying clusters that underlie incoming data by supporting the capability of reversible unsupervised learning. The potential of this work is showcased through the demonstration of successful learning in the presence of corrupted input data and probabilistic neurons, thus paving the way towards robust big-data processors
Effect of Flashlamp Heating System Parameters on the Wedge Peel Strength of Thermoplastic Carbon Fiber Tape in the Automated Tape Placement Process
No full textLaser-assisted automated tape placement systems are currently the state of the art regarding thermoplastic tape placement. Flashlamp heating systems are rather new in this field of application and offer high energy density with low safety requirements and moderate costs compared to laser-assisted automated tape placement systems. In this study, the effect of processing parameters on interlaminar bonding of carbon fiber-reinforced polyamide 6 tapes is investigated using a flashlamp heating system. The temperature during placement is monitored using an infrared camera, and the bonding strength is characterized by a wedge peel test. The bonding quality of the tapes placed between 210 °C and 330 °C at a lay-up speed of 50 mm/s is investigated. Thermogravimetric analysis, differential scanning calorimetry, and micrographs are used to investigate the material properties and effects of the processing conditions on the thermophysical properties and geometric properties of the tape. No significant changes in the thermophysical or geometric properties were found. Moisture within the tapes and staining of the quartz guides of the flashlamp system have significant influence on the bonding strength. The highest wedge peel strength of dried tapes was found at around 330 °C
Characterisation of the Mechanical Properties of Natural Fibre Polypropylene Composites Manufactured with Automated Tape Placement
No full textThe integration of natural fibre thermoplastic composites, particularly those combining flax fibres with polypropylene, offers a promising alternative to traditional synthetic composites, emphasising sustainability in composite materials. This study investigates the mechanical properties of flax/polypropylene composites manufactured using flashlamp automated tape placement and press consolidation, individually and in combination. Tensile, compression, three-point bending, and double cantilever beam tests are utilised for comparing these manufacturing processes and the mechanical performance of the resulting composites. The microstructure of the tapes is investigated using cross-sectional microscopy, and the thermophysical behaviour is analysed utilising thermogravimetric analysis and differential scanning calorimetry. The temperature during placement is monitored using an infrared camera, and the pressure is mapped with pressure-sensitive films. The natural fibre tapes show a good aptitude for being manufactured with automated tape placement. The tensile performance of tapes manufactured with automated tape placement is close to that of press consolidated samples. Compression, flexural properties, and the mode I fracture toughness critical energy release rate all benefit from a second consolidation step
Multi-area model of macaque cortex as a scaffold model and workflow testcase
No full textMulti-area model of macaque cortex as a scaffold model and workflow test caseAnno Kurth, Alexander van Meegen, Aitor Morales-Gregorio, Jari Pronold, Agnes Korcsak-Gorzo, Hannah Vollenbröker, Rembrandt Bakker, Markus Diesmann and Sacha J van AlbadaThere are many open questions on the relationships between the structure, dynamics and function of the brain, especially from a multi-modal perspective bridging micro-, meso- and macroscopic scales. Large-scale point neuron network models of cortical areas and their interconnections, integrating vast bodies of anatomical data, provide researchers with tools to investigate these issues. In order to make reliable steps in understanding, we need to take an incremental approach to the design of the models, and ensure that they can be built on by others.Here we present a multi-area model (MAM) describing all 32 areas of the macaque vision-related cortex [1] that serves as a scaffold for relating brain structure to its dynamics and function on multiple scales. The model connectivity is determined by processing available anatomical data into a layer-resolved connectome [2] of macaque vision-related cortex. A spiking neural network with the specified connectivity is constructed using NEST [3] and simulated on a supercomputer to study resting-state activity.The model is being extended and refined in various directions: In one project, the motor-related cortical areas are being added, thereby enabling the study of visuo-motor integration in a unified, biologically realistic framework. Mechanisms of spatial attention are being implemented as a first step towards modeling visual processing. Moreover, ongoing work explores the possibility of endowing the anatomically based model with information processing capabilities through learning methods for spiking neural networks [4]. The methods devised to create the macaque model are further generalized to construct a model of the human visual cortex taking into account different neuron characteristics [5] and different anatomical constraints obtained via diffusion imaging [6].Finally, a fully digitized illustrative workflow is provided alongside the MAM to ensure reproducibility and enable re-use by the community. All code is available on GitHub. The tool Snakemake [7] provides a reproducible and user-friendly framework for the execution of the model. The workflow from the anatomical data to the simulation code, analysis and visualization can serve as an example for similar data-driven brain models.[1] Schmidt, M., Bakker, R., Shen, K., Bezgin, G., Diesmann, M., & van Albada, S. J. (2018). A multi-scale layer-resolved spiking network model of resting-state dynamics in macaque visual cortical areas. PLOS Computational Biology, 14(10).[2] Schmidt, M., Bakker, R., Hilgetag, C. C., Diesmann, M., & van Albada, S. J. (2018). Multi-scale account of the network structure of macaque visual cortex. Brain Structure and Function, 223(3), 1409-1435.[3] Peyser, A. et al. (2017). NEST 2.14.0. Zenodo. 10.5281/zenodo.882971. [4] Bellec, G., Salaj, D., Subramoney, A., Legenstein, R., Maass, W. (2018). Long short-term memory and learning-to-learn in networks of spiking neurons. Advances in Neural Information Processing Systems, 787-797.[5] Teeter, C., et al. (2018). Generalized leaky integrate-and-fire models classify multiple neuron types. Nature Communications, 9, 709.[6] Van Essen, D. C., et al. (2013). The WU-Minn human connectome project: an overview. Neuroimage, 80, 62-79.[7] Köster, J., & Rahmann, S. (2012). Snakemake—a scalable bioinformatics workflow engine. Bioinformatics, 28(19), 2520-2522
Rare neural correlations implement robotic conditioning with reward delays and disturbances
No full textSoltoggio A, Lemme A, Reinhart F, Steil JJ. Rare neural correlations implement robotic conditioning with reward delays and disturbances. Frontiers in Neurorobotics. 2013;7:6.Neural conditioning associates cues and actions with following rewards. The environments in which robots operate, however, are pervaded by a variety of disturbing stimuli and uncertain timing. In particular, variable reward delays make it difficult to reconstruct which previous actions are responsible for following rewards. Such an uncertainty is handled by biological neural networks, but represents a challenge for computational models, suggesting the lack of a satisfactory theory for robotic neural conditioning. The present study demonstrates the use of rare neural correlations in making correct associations between rewards and previous cues or actions. Rare correlations are functional in selecting sparse synapses to be eligible for later weight updates if a reward occurs. The repetition of this process singles out the associating and reward-triggering pathways, and thereby copes with distal rewards. The neural network displays macro-level classical and operant conditioning, which is demonstrated in an interactive real-life human-robot interaction. The proposed mechanism models realistic conditioning in humans and animals and implements similar behaviors in neuro-robotic platforms
Reservoir computing with output feedback
No full textReinhart RF. Reservoir computing with output feedback. Bielefeld: Bielefeld University; 2011.A dynamical system approach to forward and inverse modeling is proposed. Forward and inverse models are trained in associative recurrent neural networks that are based on non-linear random projections. Feedback of estimated outputs into such reservoir networks is a key ingredient in the context of bidirectional association but entails the problem of error amplification. Robust training of reservoir networks with output feedback is achieved by a novel one-shot learning and regularization method for input-driven recurrent neural networks. It is shown that output feedback enables the implementation of ambiguous inverse models by means of multi-stable dynamics. The proposed methodology is applied to movement generation of robotic manipulators in a feedforward-feedback control framework
Neuromorphic hardware in the loop: Training a deep spiking network on the BrainScaleS wafer-scale system
No full textConcrete in the low carbon era:proceedings of the International Conference held at the University of Dundee, Scotland, UK on 9 - 11 July 2012
No full textConcrete is used globally and the components are widely available. The activity of construction is also global and both advanced and developing countries aspire to improve living conditions and infrastructure that consumes large quantities of energy and materials continuously. A different attitude to concrete, manufacture and use needs to be developed if we are to address, if not redress, the consequences of in-action. Concrete in some form or another is responsible for our civilised wellbeing. However, its manufacture and use also require substantial energy use and carbon dioxide emissions and the consequences are global. We all have a duty of care to behave responsibly and the issues of profligacy with respect to these along with related wasteful and polluting activities need to be dealt with. To do this in a committed and balanced way requires both knowledge and experience.<br/
