Searcharxiv⌕ Search

arXiv subjects

Michael Taylor

Publications and source records attributed to Michael Taylor.

At least 37 records · Page 2Linked to original sources
arXiv2021

Learning Bipedal Robot Locomotion from Human Movement

Teaching an anthropomorphic robot from human example offers the opportunity to impart humanlike qualities on its movement. In this work we present a reinforcement learning based method for teaching a real world bipedal robot to perform movements directly from human motion capture data. Our method seamlessly transitions from training in a simulation environment to executing on a physical robot without requiring any real world training iterations or offline steps. To overcome the disparity in joint configurations between the robot and the motion capture actor, our method incorporates motion re-targeting into the training process. Domain randomization techniques are used to compensate for the differences between the simulated and physical systems. We demonstrate our method on an internally developed humanoid robot with movements ranging from a dynamic walk cycle to complex balancing and waving. Our controller preserves the style imparted by the motion capture data and exhibits graceful failure modes resulting in safe operation for the robot. This work was performed for research purposes only.

cs.RO↗
arXiv2021

Pure Tensor Program Rewriting via Access Patterns (Representation Pearl)

Tensor kernels in machine learning (ML) often correspond to pure mathematical expressions, making term rewriting an attractive strategy for optimization and mapping to specialized hardware accelerators. However, existing ML intermediate representations (IRs) tend to either be \textit{pure but high-level}, making low-level rewrites to hardware targets inexpressible, or \textit{low-level but impure}, hampering the use of term rewriting altogether. This paper introduces Glenside, a pure IR whose core abstraction -- the \textit{access pattern} -- enables low-level, layout-aware, hardware-centric program rewrites. We demonstrate how term rewriting in Glenside can be used to map program fragments to hardware accelerator invocations and automatically discover classic data layout transformations like \texttt{im2col}. Glenside establishes a new foundation for exploring further term rewriting techniques in optimizing low-level tensor programs.

cs.PL↗
arXiv2021

Q-VR: System-Level Design for Future Mobile Collaborative Virtual Reality

High Quality Mobile Virtual Reality (VR) is what the incoming graphics technology era demands: users around the world, regardless of their hardware and network conditions, can all enjoy the immersive virtual experience. However, the state-of-the-art software-based mobile VR designs cannot fully satisfy the realtime performance requirements due to the highly interactive nature of user's actions and complex environmental constraints during VR execution. Inspired by the unique human visual system effects and the strong correlation between VR motion features and realtime hardware-level information, we propose Q-VR, a novel dynamic collaborative rendering solution via software-hardware co-design for enabling future low-latency high-quality mobile VR. At software-level, Q-VR provides flexible high-level tuning interface to reduce network latency while maintaining user perception. At hardware-level, Q-VR accommodates a wide spectrum of hardware and network conditions across users by effectively leveraging the computing capability of the increasingly powerful VR hardware. Extensive evaluation on real-world games demonstrates that Q-VR can achieve an average end-to-end performance speedup of 3.4x (up to 6.7x) over the traditional local rendering design in commercial VR devices, and a 4.1x frame rate improvement over the state-of-the-art static collaborative rendering.

cs.AR↗
arXiv2020

Open Access Books in the Humanities and Social Sciences: an Open Access Altmetric Advantage

The last decade has seen two significant phenomena emerge in research communication: the rise of open access (OA) publishing, and evidence of online sharing in the form of altmetrics. There has been limited examination of the effect of OA on online sharing for journal articles, and little for books. This paper examines the altmetrics of a set of 32,222 books (of which 5% are OA) and a set of 220,527 chapters (of which 7% are OA) indexed by the scholarly database Dimensions in the Social Sciences and Humanities. Both OA books and chapters have significantly higher use on social networks, higher coverage in the mass media and blogs, and evidence of higher rates of social impact in policy documents. OA chapters have higher rates of coverage on Wikipedia than their non-OA equivalents, and are more likely to be shared on Mendeley. Even within the Humanities and Social Sciences, disciplinary differences in altmetric activity are evident. The effect is confirmed for chapters, although sampling issues prevent the strong conclusion that OA facilitates extra attention at whole book level, the apparent OA altmetrics advantage suggests that the move towards OA is increasing social sharing and broader impact.

cs.DL↗
arXiv2017

The Maximal Entropy Measure of Fatou Boundaries

We look at the maximal entropy (MME) measure of the boundaries of connected components of the Fatou set of a rational map of degree greater than or equal to 2. We show that if there are infinitely many Fatou components, and if either the Julia set is disconnected or the map is hyperbolic, then there can be at most one Fatou component whose boundary has positive MME measure. We also replace hyperbolicity by the more general hypothesis of geometric finiteness.

math.DS↗
arXiv2017

Novel processes and metrics for a scientific evaluation rooted in the principles of science - Version 1

Scientific evaluation is a determinant of how scientists, institutions and funders behave, and as such is a key element in the making of science. In this article, we propose an alternative to the current norm of evaluating research with journal rank. Following a well-defined notion of scientific value, we introduce qualitative processes that can also be quantified and give rise to meaningful and easy-to-use article-level metrics. In our approach, the goal of a scientist is transformed from convincing an editorial board through a vertical process to convincing peers through an horizontal one. We argue that such an evaluation system naturally provides the incentives and logic needed to constantly promote quality, reproducibility, openness and collaboration in science. The system is legally and technically feasible and can gradually lead to the self-organized reappropriation of the scientific process by the scholarly community and its institutions. We propose an implementation of our evaluation system with the platform "the Self-Journals of Science" (www.sjscience.org).

cs.DL↗
arXiv2016

Power Side Channels in Security ICs: Hardware Countermeasures

Power side-channel attacks are a very effective cryptanalysis technique that can infer secret keys of security ICs by monitoring the power consumption. Since the emergence of practical attacks in the late 90s, they have been a major threat to many cryptographic-equipped devices including smart cards, encrypted FPGA designs, and mobile phones. Designers and manufacturers of cryptographic devices have in response developed various countermeasures for protection. Attacking methods have also evolved to counteract resistant implementations. This paper reviews foundational power analysis attack techniques and examines a variety of hardware design mitigations. The aim is to highlight exposed vulnerabilities in hardware-based countermeasures for future more secure implementations.

cs.CR↗
arXiv2015

Space-Time Reference with an Optical Link

We describe a method for realizing a high-performance Space-Time Reference (STR) using a stable atomic clock in a precisely defined orbit and synchronizing the orbiting clock to high-accuracy atomic clocks on the ground. The synchronization would be accomplished using a two-way lasercom link between ground and space. The basic concept is to take advantage of the highest-performance cold-atom atomic clocks at national standards laboratories on the ground and to transfer that performance to an orbiting clock that has good stability and that serves as a "frequency-flywheel" over time-scales of a few hours. The two-way lasercom link would also provide precise range information and thus precise orbit determination (POD). With a well-defined orbit and a synchronized clock, the satellite cold serve as a high-accuracy Space-Time Reference, providing precise time worldwide, a valuable reference frame for geodesy, and independent high-accuracy measurements of GNSS clocks. With reasonable assumptions, a practical system would be able to deliver picosecond timing worldwide and millimeter orbit determination.

astro-ph.IM↗
arXiv2015

Euler Equation on a Rotating Surface

We study 2D Euler equations on a rotating surface, subject to the effect of the Coriolis force, with an emphasis on surfaces of revolution. We bring in conservation laws that yield long time estimates on solutions to the Euler equation, and examine ways in which the solutions behave like zonal fields, building on work of B.~Cheng and A.~Mahalov, examining how such 2D Euler equations can account for the observed band structure of rapidly rotating planets. Specific results include both an analysis of time averages of solutions and a study of stability of stationary zonal fields. The latter study includes both analytical and numerical work.

math.AP↗
arXiv2014

A comparative analysis of numerical approaches to the mechanics of elastic sheets

Numerically simulating deformations in thin elastic sheets is a challenging problem in computational mechanics due to destabilizing compressive stresses that result in wrinkling. Determining the location, structure, and evolution of wrinkles in these problems have important implications in design and is an area of increasing interest in the fields of physics and engineering. In this work, several numerical approaches previously proposed to model equilibrium deformations in thin elastic sheets are compared. These include standard finite element-based static post-buckling approaches as well as a recently proposed method based on dynamic relaxation, which are applied to the problem of an annular sheet with opposed tractions where wrinkling is a key feature. Numerical solutions are compared to analytic predictions, enabling a quantitative evaluation of the predictive power of the various methods. Results indicate that static finite element approaches are highly sensitive to initial imperfections, relying on \textit{a priori} knowledge of the equilibrium wrinkling pattern to generate optimal results. In contrast, dynamic relaxation is much less sensitive to initial imperfections and can generate solutions for a wide variety of loading conditions without requiring knowledge of the equilibrium solution beforehand.

cond-mat.mtrl-sci↗
arXiv2012

Nonlinear bound states on weakly homogeneous spaces

We prove the existence of ground state solutions for a class of nonlinear elliptic equations, arising in the production of standing wave solutions to an associated family of nonlinear Schrödinger equations. We examine two constrained minimization problems, which give rise to such solutions. One yields what we call $F_λ$-minimizers, the other energy minimizers. We produce such ground state solutions on a class of Riemannian manifolds called weakly homogeneous spaces, and establish smoothness, positivity, and decay properties. We also identify classes of Riemannian manifolds with no such minimizers, and classes for which essential uniqueness of positive solutions to the associated elliptic PDE fails.

math.AP↗
arXiv2011

Epidemic threshold and control in a dynamic network

In this paper we present a model describing Susceptible-Infected-Susceptible (SIS) type epidemics spreading on a dynamic contact network with random link activation and deletion where link ac- tivation can be locally constrained. We use and adapt a improved effective degree compartmental modelling framework recently proposed by Lindquist et al. [J. Lindquist et al., J. Math Biol. 62, 2, 143 (2010)] and Marceau et al. [V. Marceau et al., Phys. Rev. E 82, 036116 (2010)]. The resulting set of ordinary differential equations (ODEs) is solved numerically and results are compared to those obtained using individual-based stochastic network simulation. We show that the ODEs display excellent agreement with simulation for the evolution of both the disease and the network, and is able to accurately capture the epidemic threshold for a wide range of parameters. We also present an analytical R0 calculation for the dynamic network model and show that depending on the relative timescales of the network evolution and disease transmission two limiting cases are recovered: (i) the static network case when network evolution is slow and (ii) homogeneous random mixing when the network evolution is rapid. We also use our threshold calculation to highlight the dangers of relying on local stability analysis when predicting epidemic outbreaks on evolving networks.

math.DS↗
arXiv2007

100 Years of dimensional analysis: New steps toward empirical law deduction

On the verge of the centenary of dimensional analysis (DA), we present a generalisation of the theory and a methodology for the discovery of empirical laws from observational data. It is well known that DA: a) reduces the number of free parameters, b) guarantees scale invariance through dimensional homogeneity and c) extracts functional information encoded in the dimensionless grouping of variables. Less known are the results of Rudolph and co-workers that DA also gives rise to a new pair of transforms - the similarity transform (S) that converts physical dimensional data into dimensionless space and its inverse (S'). Here, we present a new matrix generalisation of the Buckingham Theorem, made possible by recent developments in the theory of inverse non-square matrices, and show how the transform pair arises naturally. We demonstrate that the inverse transform S' is non-unique and how this casts doubt on scaling relations obtained in cases where observational data has not been referred to in order to break the degeneracy inherent in transforming back to dimensional (physical) space. As an example, we show how the underlying functional form of the Planck Radiation Law can be deduced in only a few lines using the matrix method and without appealing to first principles; thus demonstrating the possibility of a priori knowledge discovery; but that subsequent data analysis is still required in order to identify the exact causal law. It is hoped that the proof presented here will give theoreticians confidence to pursue inverse problems in physics using DA.

physics.class-ph↗
arXiv2007

Integrability of Rough Almost Complex Structures

We extend the Newlander-Nirenberg theorem to manifolds with almost complex structures that have somewhat less than Lipschitz regularity. We also discuss the regularity of local holomorphic coordinates in the integrable case, with particular attention to Lipschitz almost complex structures.

math.DG↗
arXiv2007

The Complex Frobenius Theorem for Rough Involutive Structures

We establish a version of the complex Frobenius theorem in the context of a complex subbundle S of the complexified tangent bundle of a manifold, having minimal regularity. If the subbundle S defines the structure of a Levi-flat CR-manifold, it suffices that S be Lipschitz for our results to apply. A principal tool in the analysis is a precise version of the Newlander-Nirenberg theorem with parameters, for integrable almost complex structures with minimal regularity, which builds on previous recent work of the authors.

math.DG↗
arXiv2007

Exact solutions for the populations of the n-level ion

We present a matrix solution to the full equations of statistical equilibrium that give the energy level populations of collisionally-excited ions in photoionised gaseous nebulae. The rationale for such a calculation is to maintain a parity between improvements in the quantum-mechanically evaluated values for collision strengths and transition probabilities from the Iron and Opacity Projects on the one hand, and 3D photoionisation codes such as MOCASSIN and astrophysical software for producing nebular diagnostics such as the Nebular package for IRAF, on the other. We have taken advantage of the fact that mathematics programs such as MATLAB and Mathematica have proven to be very adept at symbolic manipulation providing a route to exact solutions for the n-level ion. In particular, we have avoided the substitution of estimated values. We provide the matrix solution for the 5-level ion as an example and show how the equations faithfully reduce to the exact solution for the 3-level ion. Through the forbidden line ratio R23, we compare the exact solution with a) that obtained from the observed emission of the spherical planetary nebula Abell 39, b) 3D Monte-Carlo photoionisation modelling of the same nebula, c) the approximate 5-level program TEMDEN and d) the exact 3-level ion. The general solution presented here means that programs for the calculation of level populations can obtain solutions for ions with a user-specified number of excited levels. The use of a separate and updatable database of atomic and ionic constants such as that provided by NIST, means that software of more general application can now be made available; particularly for the study of high excitation objects such as active galactic nebulae (AGNs) and supernovae (SNs) where higher excited levels become significant.

astro-ph↗
arXiv2007

On the deduction of galaxy abundances with evolutionary neural networks

A growing number of indicators are now being used with some confidence to measure the metallicity(Z) of photoionisation regions in planetary nebulae, galactic HII regions(GHIIRs), extra-galactic HII regions(EGHIIRs) and HII galaxies(HIIGs). However, a universal indicator valid also at high metallicities has yet to be found. Here, we report on a new artificial intelligence-based approach to determine metallicity indicators that shows promise for the provision of improved empirical fits. The method hinges on the application of an evolutionary neural network to observational emission line data. The network's DNA, encoded in its architecture, weights and neuron transfer functions, is evolved using a genetic algorithm. Furthermore, selection, operating on a set of 10 distinct neuron transfer functions, means that the empirical relation encoded in the network solution architecture is in functional rather than numerical form. Thus the network solutions provide an equation for the metallicity in terms of line ratios without a priori assumptions. Tapping into the mathematical power offered by this approach, we applied the network to detailed observations of both nebula and auroral emission lines in the optical for a sample of 96 HII-type regions and we were able to obtain an empirical relation between Z and S23 with a dispersion of only 0.16 dex. We show how the method can be used to identify new diagnostics as well as the nonlinear relationship supposed to exist between the metallicity Z, ionisation parameter U and effective (or equivalent) temperature T*.

astro-ph↗