arXiv subjects

Richard Lau

Publications and source records attributed to Richard Lau.

9 recordsLinked to original sources

arXiv2024

Temporal-Spatial Processing of Event Camera Data via Delay-Loop Reservoir Neural Network

This paper describes a temporal-spatial model for video processing with special applications to processing event camera videos. We propose to study a conjecture motivated by our previous study of video processing with delay loop reservoir (DLR) neural network, which we call Temporal-Spatial Conjecture (TSC). The TSC postulates that there is significant information content carried in the temporal representation of a video signal and that machine learning algorithms would benefit from separate optimization of the spatial and temporal components for intelligent processing. To verify or refute the TSC, we propose a Visual Markov Model (VMM) which decompose the video into spatial and temporal components and estimate the mutual information (MI) of these components. Since computation of video mutual information is complex and time consuming, we use a Mutual Information Neural Network to estimate the bounds of the mutual information. Our result shows that the temporal component carries significant MI compared to that of the spatial component. This finding has often been overlooked in neural network literature. In this paper, we will exploit this new finding to guide our design of a delay-loop reservoir neural network for event camera classification, which results in a 18% improvement on classification accuracy.

cs.CV↗

arXiv2021

Scaled-Time-Attention Robust Edge Network

This paper describes a systematic approach towards building a new family of neural networks based on a delay-loop version of a reservoir neural network. The resulting architecture, called Scaled-Time-Attention Robust Edge (STARE) network, exploits hyper dimensional space and non-multiply-and-add computation to achieve a simpler architecture, which has shallow layers, is simple to train, and is better suited for Edge applications, such as Internet of Things (IoT), over traditional deep neural networks. STARE incorporates new AI concepts such as Attention and Context, and is best suited for temporal feature extraction and classification. We demonstrate that STARE is applicable to a variety of applications with improved performance and lower implementation complexity. In particular, we showed a novel way of applying a dual-loop configuration to detection and identification of drone vs bird in a counter Unmanned Air Systems (UAS) detection application by exploiting both spatial (video frame) and temporal (trajectory) information. We also demonstrated that the STARE performance approaches that of a State-of-the-Art deep neural network in classifying RF modulations, and outperforms Long Short-term Memory (LSTM) in a special case of Mackey Glass time series prediction. To demonstrate hardware efficiency, we designed and developed an FPGA implementation of the STARE algorithm to demonstrate its low-power and high-throughput operations. In addition, we illustrate an efficient structure for integrating a massively parallel implementation of the STARE algorithm for ASIC implementation.

cs.CV↗

arXiv2017

SO(N) gauge theories in 2+1 dimensions: glueball spectra and confinement

We calculate the spectrum of light glueballs and the string tension in a number of SO(N) lattice gauge theories in 2+1 dimensions, with N in the range from N=3 to N=16. After extrapolating to the continuum limit and then to N=oo we compare to the spectrum and string tension of the SU(N=oo) gauge theory and find that the most reliably and precisely calculated physical quantities are consistent in that limit. We also compare the glueball spectra of those pairs of SO(N) and SU(N') theories that possess the same Lie algebra, i.e. SO(3) and SU(2), SO(4) and SU(2)xSU(2), SO(6) and SU(4), and find that for the very lightest glueballs the spectra are consistent within each such pair, as are the string tensions and the couplings. Where there are apparent discrepancies they are typically for heavier glueballs, where the systematic errors are much harder to control. We calculate the SO(N) string tensions with a particular focus on the confining properties of SO(2N+1) theories which, unlike SO(2N) theories, possess a trivial centre. We find that for both the light glueballs and for the string tension SO(2N) and SO(2N+1) gauge theories appear to form a single smooth sequence.

hep-lat↗

arXiv2016

The deconfining phase transition of SO(N) gauge theories in 2+1 dimensions

We calculate the deconfining temperature of SO(N) gauge theories in 2+1 dimensions, and determine the order of the phase transition as a function of N, for various values of N in the range [4,16]. We do so by extrapolating our lattice results to the infinite volume limit, and then to the continuum limit, for each value of N. We then extrapolate to the N=infinity limit and observe that the SO(N) and SU(N) deconfining temperatures agree in that limit. We find that the the deconfining temperatures of all the SO(N) gauge theories appear to follow a single smooth function of N, despite the lack of a non-trivial centre for odd N. We also compare the deconfining temperatures of SO(6) with SU(4), and of SO(4) with SU(2)xSU(2), motivated by the fact that these pairs of gauge theories share the same Lie algebras.

hep-lat↗

arXiv2016

On the weak N-dependence of SO(N) and SU(N) gauge theories in 2+1 dimensions

We consider (continuum) mass ratios of the lightest `glueballs' as a function of N for SO(N) and SU(N) lattice gauge theories in D=2+1. We observe that the leading large N correction is usually sufficient to describe the N-dependence of SO(N.geq.3) and SU(N.geq.2), within the errors of the numerical calculation. Just as interesting is the fact that the coefficient of this correction almost invariably turns out to be anomalously small, for both SO(N) and SU(N). We point out that this can follow naturally from the strong constraints that one naively expects from the Lie algebra equivalence between certain SO(N) and SU(N') theories and the equivalence of SO(infinity) and SU(infinity). The same argument for a weak N-dependence can in principle apply to SU(N) and SO(N) gauge theories in D=3+1.

hep-lat↗

arXiv2015

Deconfining temperatures in SO(N) and SU(N) gauge theories

We present our current results for the deconfining temperatures in SO(N) gauge theories in 2+1 dimensions. SO(2N) theories may help us to understand QCD at finite chemical potential since there is a large-N orbifold equivalence between SO(2N) QCD-like theories and SU(N) QCD, and SO(2N) theories do not have the sign problem present in QCD. We show that the deconfining temperatures in these two theories match at the large-N limit. We also present results for SO(2N+1) gauge theories and compare results for SO(6) with SU(4) gauge theories, which have the same Lie algebras but different centres.

hep-lat↗

arXiv2013

SO(2N) and SU(N) gauge theories

We present our preliminary results of SO(2N) gauge theories, approaching the large-N limit. SO(2N) theories may help us to understand QCD at finite chemical potential since there is an orbifold equivalence between SO(2N) and SU(N) gauge theories at large-N and SO(2N) theories do not have the sign problem present in QCD. We consider the string tensions, mass spectra, and deconfinement temperatures in the SO(2N) pure gauge theories in 2+1 dimensions, comparing them to their corresponding SU(N) theories.

hep-lat↗

arXiv2012

SO(2N) and SU(N) gauge theories in 2+1 dimensions

We perform an exploratory investigation of how rapidly the physics of SO(2N) gauge theories approaches its N=oo limit. This question has recently become topical because SO(2N) gauge theories are orbifold equivalent to SU(N) gauge theories, but do not have a finite chemical potential sign problem. We consider only the pure gauge theory and, because of the inconvenient location of the lattice strong-to-weak coupling 'bulk' transition in 3+1 dimensions, we largely confine our numerical calculations to 2+1 dimensions. We discuss analytic expectations in both D=2+1 and D=3+1, show that the SO(6) and SU(4) spectra do indeed appear to be the same, and show that a number of mass ratios do indeed appear to agree in the large-N limit. In particular SO(6) and SU(3) gauge theories are quite similar except for the values of the string tension and coupling, both of which differences can be readily understood.

hep-lat↗

arXiv2012

On Quantification of Anchor Placement

This paper attempts to answer a question: for a given traversal area, how to quantify the geometric impact of anchor placement on localization performance. We present a theoretical framework for quantifying the anchor placement impact. An experimental study, as well as the field test using a UWB ranging technology, is presented. These experimental results validate the theoretical analysis. As a byproduct, we propose a two-phase localization method (TPLM) and show that TPLM outperforms the least-square method in localization accuracy by a huge margin. TPLM performs much faster than the gradient descent method and slightly better than the gradient descent method in localization accuracy. Our field test suggests that TPLM is more robust against noise than the least-square and gradient descent methods.

cs.NI↗