Filters








15 Hits in 7.0 sec

Training deep neural networks for binary communication with the Whetstone method

William Severa, Craig M. Vineyard, Ryan Dellana, Stephen J. Verzi, James B. Aimone
2019 Nature Machine Intelligence  
We present a method for training deep spiking neural networks using an iterative modification of the backpropagation optimization algorithm.  ...  This paper presents a new technique for training networks for low-precision communication.  ...  Overview of Whetstone Process. Whetstone is a process for training binary, threshold-activation spiking neural networks using existing deep learning methods.  ... 
doi:10.1038/s42256-018-0015-y fatcat:pcoqu5ekibbcjj4q2mropjjlfe

Hyperparameter Optimization in Binary Communication Networks for Neuromorphic Deployment [article]

Maryam Parsa, Catherine D. Schuman, Prasanna Date, Derek C. Rose, Bill Kay, J. Parker Mitchell, Steven R. Young, Ryan Dellana, William Severa, Thomas E. Potok, Kaushik Roy
2020 arXiv   pre-print
In this work, we introduce a Bayesian approach for optimizing the hyperparameters of an algorithm for training binary communication networks that can be deployed to neuromorphic hardware.  ...  Training neural networks for neuromorphic deployment is non-trivial.  ...  from traditional artificial neural network training.  ... 
arXiv:2005.04171v1 fatcat:rovwd5e24jhyvabh7pf2hobrzq

A Survey on Spiking Neural Networks

Chan Sik Han, Keon Myung Lee
2021 International Journal of Fuzzy Logic and Intelligent Systems  
Spiking neural networks (SNNs) have attracted attention as the third generation of neural networks for their promising characteristics of energy-efficiency and biological plausibility.  ...  This paper provides a gentle survey of SNNs to give an overview of what they are and how they are trained.  ...  Spiking neural networks (SNNs) are computation models which mimic biological neural networks in a more similar way than artificial neural networks (ANNs) [6] .  ... 
doi:10.5391/ijfis.2021.21.4.317 fatcat:id42nlnfvrf4bcyudgoyyuebg4

Machine Learning and Big Scientific Data [article]

Tony Hey, Keith Butler, Sam Jackson, Jeyarajan Thiyagalingam
2019 arXiv   pre-print
Remarkably, they have been able to achieve some spectacular results for this specific scientific problem. Can deep learning be similarly transformative for other scientific problems?  ...  Google's DeepMind has now also used deep learning technology to develop their AlphaFold tool to make predictions for protein folding.  ...  For this reason, instead of using a more flexible model like a convolutional neural network (CNN) and deep learning, for this sub-benchmark we use a simple, multi-layer neural network for the baseline  ... 
arXiv:1910.07631v1 fatcat:fpwolsvxmbc7dci3lzwv3r3ba4

Advancing Neuromorphic Computing With Loihi: A Survey of Results and Outlook

Mike Davies, Andreas Wild, Garrick Orchard, Yulia Sandamirskaya, Gabriel A. Fonseca Guerra, Prasad Joshi, Philipp Plank, Sumedh R. Risbud
2021 Proceedings of the IEEE  
ABSTRACT | Deep artificial neural networks apply principles of the brain's information processing that led to breakthroughs in machine learning spanning many problem domains.  ...  While conventional feedforward deep neural networks show modest if any benefit on Loihi, more brain-inspired networks using recurrence, precise spike-timing relationships, synaptic plasticity, stochasticity  ...  The basic deep learning paradigm of applying the error backpropagation algorithm to differentiable artificial neural networks (ANNs) has proven to be a powerful tool for optimizing these high-dimensional  ... 
doi:10.1109/jproc.2021.3067593 fatcat:krqdmy3u6jdvfl7btjglek5ag4

Bayesian Multi-objective Hyperparameter Optimization for Accurate, Fast, and Efficient Neural Network Accelerator Design

Maryam Parsa, John P. Mitchell, Catherine D. Schuman, Robert M. Patton, Thomas E. Potok, Kaushik Roy
2020 Frontiers in Neuroscience  
In addition, we demonstrate resiliency of different input/output encoding, training neural network, or the underlying accelerator modules in a neuromorphic system to the changes of the hyperparameters.  ...  A key challenge is to find the optimum set of hyperparameters that might belong to the input/output encoding modules, the neural network itself, the application, or the underlying hardware.  ...  For experiments on Artificial Neural Networks (ANNs), we select PUMA (Ankit et al., 2019) as the underlying hardware with two different deep neural network architectures, AlexNet (Krizhevsky et al.,  ... 
doi:10.3389/fnins.2020.00667 pmid:32848531 pmcid:PMC7396641 fatcat:jud4jgv3ejawjjtfcxeqtzza5a

High Performance Computing Environment using General Purpose Computations on Graphics Processing Unit

Andreas Widjaja, Tjatur Kandaga Gautama, Sendy Ferdian Sujadi, Steven Rumanto Harnandy
2021 Jurnal Teknik Informatika dan Sistem Informasi  
The goal of this paper is to show a design of a HPC which is capable of running complex and multi-threaded computations.  ...  For starters, the HPC environment will be served for computational projects of students and members of the Faculty of Information Technology, Universitas Kristen Maranatha.  ...  Lembaga Penelitian dan Pengabdian Kepada Masyarakat (LPPM) Universitas Kristen Maranatha is also acknowledged for facilitating the administrative process necessary for this research.  ... 
doi:10.28932/jutisi.v7i2.3715 fatcat:rzr2vtnplfdpplugnhgamdo5cm

D1.1 - State of the Art Analysis

Danilo Ardagna
2021 Zenodo  
Then, the deliverable provides a background on AI applications design, also considering some advanced design trends (e.g., Network Architecture Search, Federated Learning, Deep Neural Networks partitioning  ...  It examines the standardisation and opensource landscape, where many of the emerging standards are mainly focused on the network layer, also highlighting opportunities for open-source community involvement  ...  In particular, we are interested in the Artificial Intelligence modules: • NVIDIA TensorRT: provides a high-performance neural network inference engine for production deployment of deep learning applications  ... 
doi:10.5281/zenodo.6372377 fatcat:f6ldfuwivbcltew4smiiwphfty

Proceedings of the 5th bwHPC Symposium [article]

Michael Janczyk, Dirk Von Suchodoletz, Bernd Wiebelt, University, My, University, My
2019
Additionally, the symposium served as a platform for discussing and exchanging ideas concerning the use of these large scientific infrastructures as well as its further development.  ...  The bwHPC initiative, founded by the Ministry of Science, Research and the Arts and the universities in Baden-Württemberg, is a state-wide federated approach aimed at assisting scientists with mastering  ...  Rousseau and A. U. J. Lode for their help regarding the implementation of the SGF algorithm and Mctdhx method, respectively. We also thank T.  ... 
doi:10.15496/publikation-29062 fatcat:rtffnd5u6zeoxaezoijeyv4ldy

Video-based Bed Monitoring

Manuel Martinez
2017
We also present a method for processing depth video targeted to our scenario named Bed Aligned Maps (BAM).  ...  The datasets are large enough to train machine learning methods and obtain statistically significant results.  ...  We hope that our contributions will help promote a wider adaption EMD as a loss criterion within deep learning frameworks.  ... 
doi:10.5445/ir/1000076972 fatcat:k4wggdg3sngavnv262s5fyicnu

Oral Presentations 2022 AANS Annual Scientific Meeting

Philadelphia, Pennsylvania • April 29–May 2, 2022
2022
Future studies are warranted to further establish safety, efficacy, and support for evEEG as a potential tool for neural recording, deep brain stimulation, and brain-machine-interface.  ...  Endovascular electroencephalography (evEEG) utilizes the cerebrovascular system as a minimally-invasive conduit to record electrical activity from adjacent neural structures, mitigating the poor spatial  ...  Conclusion: We identified T cell clonal expansion in the blood as a biomarker of response to ICI.  ... 
doi:10.3171/2022.5.jns.aans2022abstracts pmid:35535820 fatcat:3aijbxn7r5b3pkccuqnpf3tieq

The clinical utility of the assessment of learning potential following brain injury [article]

Stephanie Margaret Uprichard, UH Research Archive, UH Research Archive
2014
In this research dynamic testing involved a pre and post test administration that sandwiched a training element.  ...  Learning potential measures a latent or dormant ability that is brought out by a third party during dynamic training.  ...  Errorless learning for this research is seen as a useful tool to overcome memory problems often encountered following an ABI. It should not necessarily be seen as a method for all models of training.  ... 
doi:10.18745/th.14354 fatcat:egplnjlt6zdwfkfzdnz7kbrjye

OASIcs, Volume 45, CMN'15, Complete Volume [article]

Mark A. Finlayson, Ben Miller, Antonio Lieto, Remi Ronfard
2015
For further information, see http://ehumanities.nl. Mike Kestemont has been supported for this work as a postdoctoral researcher for the Research Foundation Flanders (FWO).  ...  We would also like to thank Brandon Tearse for the development of Skald, Peter Mawhorter for his assistance in reimplementation, and Noah Wardrip-Fruin for his feedback on SIG representation.  ...  neural network framework [28] , EST has yet to be applied in a symbolic architecture.  ... 
doi:10.4230/oasics.cmn.2015 fatcat:mwish2iha5eodnlbiymk27qjvi

Hate or glory: a categorical and experimental consideration of Bronze Age halberds in Scotland in relation to MBA weaponry [article]

Rachel Faulkner-Jones, University Of Edinburgh, Ian Ralston, Manuel Fernandez-Gotz
2021
or mobility, and the mending and conservation evidence in the prehistoric assemblage is hypothesised to be linked to their role as combat and political power proxies in long-distance communication networks  ...  tissue proxy, and secondly a pig carcass as a soft tissue proxy.  ...  a wide network of trade and communication.  ... 
doi:10.7488/era/1158 fatcat:zg5ie773zjh6tjpofgb24abxpi

Effects of Movement on Biometric Facial Recognition in Body-Worn Cameras

Julia Bryan
2020
The general conclusion of this study is that a body-worn camera is not a suitable sensor for a biometric facial recognition system at this time, though advances in camera technology and biometric systems  ...  In the second phase, the researcher collected quantitative data using a single facial recognition subject and a static body-worn camera mounted to an adjustable tripod.  ...  FaceID employs what it calls the TrueDepth camera system and unspecified algorithms to "accurately map the geometry of [the operator's] face," and then applies "neural networks for matching and anti-spoofing  ... 
doi:10.25394/pgs.12227372 fatcat:mqc4fwkkbvhd3k5dshxxjxkfgm