A5014926556- Neural Networks and Applications
- Visual perception and processing mechanisms
- Galaxies: Formation, Evolution, Phenomena
- Handwritten Text Recognition Techniques
- Gamma-ray bursts and supernovae
- Advanced Memory and Neural Computing
- Stellar, planetary, and galactic studies
- Image Processing and 3D Reconstruction
- Neural Networks and Reservoir Computing
- Hand Gesture Recognition Systems
- Adaptive optics and wavefront sensing
- Opinion Dynamics and Social Influence
- Online Learning and Analytics
- Natural Language Processing Techniques
- Gaussian Processes and Bayesian Inference
- Astronomy and Astrophysical Research
- Complex Network Analysis Techniques
- Color perception and design
- Color Science and Applications
- Ocular and Laser Science Research
- ICT in Developing Communities
- Sensor Technology and Measurement Systems
- Horticultural and Viticultural Research
- Statistical and numerical algorithms
- Agricultural Innovations and Practices
The Open University
2017-2023
The University of Western Australia
2022
Association for Computing Machinery
2020
Naval Postgraduate School
2020
Hong Kong Metropolitan University
2014
University of Cambridge
2004
Institute of Astronomy
2004
Nokia (United States)
1991-2002
AT&T (United States)
1992-1994
Imperial College London
1986
This paper describes the development of an algorithm for verification signatures written on a touch-sensitive pad. The signature is based artificial neural network. novel network presented here, called “Siamese” time delay network, consists two identical networks joined at their output. During training learns to measure similarity between pairs signatures. When used verification, only one half Siamese evaluated. output this feature vector input signature. Verification comparing with stored...
The architecture, implementation, and applications of a special-purpose neural network processor are described. chip performs over 2000 multiplications additions simultaneously. Its data path is particularly suitable for the convolutional topologies that typical in classification networks, but can also be configured fully connected or feedback topologies. Resources multiplexed to permit implementation networks with several hundreds thousands connections on single chip. Computations performed...
A neural network with 136000 connections for recognition of handwritten digits has been implemented using a mixed analog/digital chip. The chip is capable processing 1000 characters/s. system essentially the same rate (5%) as simulation 32-b floating-point precision.
The Euclid telescope, due for launch in 2021, will perform an imaging and slitless spectroscopy survey over half the sky, to map baryon wiggles weak lensing. During is expected resolve 100,000 strong gravitational lens systems. This ideal find rare configurations, provided they can be identified reliably on a reasonable timescale. For this reason we have developed Convolutional Neural Network (CNN) that used identify images containing lensing CNNs already been image digit classification as...
ABSTRACT We investigate the ability of human ‘expert’ classifiers to identify strong gravitational lens candidates in Dark Energy Survey like imaging. recruited a total 55 people that completed more than 25 per cent project. During classification task, we present participants 1489 images. The sample contains variety data including simulations, real lenses, non-lens examples, and unlabelled data. find experts are extremely good at finding bright, well-resolved Einstein rings, while arcs with...
ABSTRACT Forthcoming large imaging surveys such as Euclid and the Vera Rubin Observatory Legacy Survey of Space Time are expected to find more than 105 strong gravitational lens systems, including many rare exotic populations compound lenses, but these systems will be interspersed among much larger catalogues ∼109 galaxies. This volume data is too for visual inspection by volunteers alone feasible lenses only appear in a small fraction which could cause amount false positives. Machine...
A neural network algorithm-based system that reads handwritten ZIP codes appearing on real US mail is described. The uses a recognition-based segmenter, hybrid of connected-components analysis (CCA), vertical cuts, and recognizer. Connected components are single digits handled by CCA. CCs combined or dissected the vertical-cut segmenter. four main stages processing preprocessing, in which noise removed deslanted, CCA segmentation recognition, vertical-cut-point estimation segmentation,...
We investigate the constraints on models of supermassive black hole (SMBH) and quasar formation obtainable from two recent observational developments: discovery luminous quasars at z~6, estimates local mass density SMBHs. If ~90 per cent this was accreted redshifts z<3, as suggested by observed luminosity functions, these joint pose a challenge for models, which must account population z~6 within very limited `mass budget'. class based hierarchical structure scenario, in major mergers...
Education is a major force for economic and social wellbeing. Despite high aspirations, education at all levels can be expensive ineffective. Three Grand Challenges are identified: (1) enable people to learn orders of magnitude more effectively, (2) less cost, (3) demonstrate success by exemplary interdisciplinary in complex systems science. A ten year ‘man-on-the-moon’ project proposed which FuturICT’s unique combination Complexity, Social Computing Sciences could provide an urgently needed...
Wide-field sub-millimetre surveys have driven many major advances in galaxy evolution the past decade, but without extensive follow-up observations coarse angular resolution of these limits science exploitation. This has development various analytical deconvolution methods. In last half a decade Generative Adversarial Networks been used to attempt deconvolutions on optical data. Here we present an autoencoder with novel loss function overcome this problem sub-millimeter wavelength range....
A special-purpose chip, optimized for computational needs of neural networks and performing over 2000 multiplications additions simultaneously, is described. Its data path particularly suitable the convolutional architectures typical in pattern classification but can also be configured fully connected or feedback topologies. development system permits rapid prototyping new applications analysis impact specialized hardware on performance. The power flexibility processor are demonstrated with...
May 01 1993 Improving Rejection Performance on Handwritten Digits by Training with “Rubbish” In Special Collection: CogNet Jane Bromley, Bromley AT&T Bell Laboratories, Holmdel, NJ 07733, USA Search for other works this author on: This Site Google Scholar John S. Denker Author and Article Information Received: July 27 1992 Accepted: October 20 Online ISSN: 1530-888X Print 0899-7667 © Massachusetts Institute of Technology1993 Neural Computation (1993) 5 (3): 367–370....
A high-speed programmable neural network chip and its application to character recognition are described. with over 130000 connections has been implemented on a single operates at rate of 1000 classifications per second. The performs up 2000 multiplications additions simultaneously. Its datapath is suitable for the convolutional architectures that typical in pattern classification networks, but can also be configured fully connected or feedback topologies. Computations were performed 6 bits...
Journal Article A STUDY OF SYSTEMATIC VISUAL PERSEVERATION INVOLVING CENTRAL MECHANISMS Get access I. M. BLYTHE, BLYTHE Search for other works by this author on: Oxford Academic PubMed Google Scholar J. BROMLEY, BROMLEY K. H. RUDDOCK, RUDDOCK Reprint requests and correspondence to: Dr Ruddock, Department of Physics (Biophysics), Imperial College, London SW7 2BZ. C. KENNARD, KENNARD TRAUB Brain, Volume 109, Issue 4, August 1986, Pages 661–675, https://doi.org/10.1093/brain/109.4.661...
The authors outline OCR (optical character recognition) technology developed at AT&T Bell Laboratories, including a recognition network that learns feature extraction kernels and custom VLSI chip is designed for neural-net image processing. It concluded both high speed accuracy can be obtained using methods recognition. Networks learn their own kernels. Special-purpose chips combined with digital signal processors quickly evaluate character-recognition neural nets. This particularly useful...
Psychophysical methods developed for the investigation of spatial and temporal pathways in human vision have been applied combination with two-colour increment threshold technique W. S. Stiles to study way which signals from blue-sensitive cones are transmitted along visual pathways. A flicker sensitive spatio-temporal filter, designated 'ST2', has examined by background modulation methods, filters bars a specific width grating adaptation employing dichoptic presentation stimuli. It is shown...