Source author record

John Taylor

John Taylor appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

math.DS astro-ph.HE Machine Learning physics.ao-ph physics.flu-dyn

Catalog footprint

What is connected

5works

5topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

Optimizing the optimizer for data driven deep neural networks and physics informed neural networks

We investigate the role of the optimizer in determining the quality of the model fit for neural networks with a small to medium number of parameters. We study the performance of Adam, an algorithm for first-order gradient-based optimization that uses adaptive momentum, the Levenberg and Marquardt (LM) algorithm a second order method, Broyden,Fletcher,Goldfarb and Shanno algorithm (BFGS) a second order method and LBFGS, a low memory version of BFGS. Using these optimizers we fit the function y = sinc(10x) using a neural network with a few parameters. This function has a variable amplitude and a constant frequency. We observe that the higher amplitude components of the function are fitted first and the Adam, BFGS and LBFGS struggle to fit the lower amplitude components of the function. We also solve the Burgers equation using a physics informed neural network(PINN) with the BFGS and LM optimizers. For our example problems with a small to medium number of weights, we find that the LM algorithm is able to rapidly converge to machine precision offering significant benefits over other optimizers. We further investigated the Adam optimizer with a range of models and found that Adam optimiser requires much deeper models with large numbers of hidden units containing up to 26x more parameters, in order to achieve a model fit close that achieved by the LM optimizer. The LM optimizer results illustrate that it may be possible build models with far fewer parameters. We have implemented all our methods in Keras and TensorFlow 2.

preprint2021arXiv

Anthropogenic Mixing of Seasonally Stratified Shelf Seas by Offshore Wind Farm Infrastructure

The offshore wind energy sector has rapidly expanded over the past two decades, providing a renewable energy solution for coastal nations. Sector development has been led in Europe, but is growing globally. Most developments to date have been in well-mixed, i.e. unstratified, shallow-waters near to shore. Sector growth is, for the first time, pushing developments to deep water, into a brand new environment: seasonally stratified shelf seas. Seasonally stratified shelf seas, where water density varies with depth, have a disproportionately key role in primary production, marine ecosystem and biochemically cycles. Infrastructure will directly mix stratified shelf seas. The magnitude of this mixing, additional to natural background processes, has yet to be fully quantified. If large enough it may erode shelf sea stratification. Therefore, offshore wind growth may destabilize and fundamentally change shelf sea systems. However, enhanced mixing may also positively impact some marine ecosystems. This paper sets the scene for sector development into this new environment, reviews the potential physical and environmental benefits and impacts of large scale industrialization of seasonally stratified shelf seas and identifies areas where research is required to best utilise, manage and mitigate environmental change.

preprint2013arXiv

Chaos in Topological Spaces

We give a definition of chaos for a continuous self-map of a general topological space. This definition coincides with the Devanney definition for chaos when the topological space happens to be a metric space. We show that in a uniform Hausdorff space, there is a meaningful definition of sensitive dependence on initial conditions, and prove that if a map is chaotic on a such a space, then it necessarily has sensitive dependence on initial conditions. The proof is interesting in that it explains very clearly what causes a chaotic process to have sensitive dependence. Finally, we construct a chaotic map on a non-metrizable topological space.

preprint2012arXiv

Updated Results from the RICE Experiment and Future Prospects for Ultra-High Energy Neutrino Detection at the South Pole

The RICE experiment seeks observation of ultra-high energy (UHE; >100 PeV) neutrinos interacting in Antarctic ice, by measurement of the radiowavelength Cherenkov radiation resulting from the collision of a neutrino with an ice molecule. RICE was initiated in 1999 as a first-generation prototype for an eventual, large-scale in-ice UHE neutrino detector. Herein, we present updated limits on the diffuse UHE neutrino flux, based on twelve years of data taken between 1999 and 2010. We find no convincing neutrino candidates, resulting in 95% confidence-level model-dependent limits on the flux which improve on our previous (2006) results by a factor of two. We also report on ice properties' studies, including observation of radiowave reflections from layers internal to the ice sheet. We observe correlations of these reflections with the local ice flow direction, and also qualitatively verify the expected reduction in radio frequency ice transparency with increasing depth/temperature. A new experimental initiative (the Askaryan Radio Array, ARA) commenced at South Pole in 2010-11 will improve the RICE sensitivity (per unit time) by approximately two orders of magnitude.

preprint2007arXiv

When will a One Parameter Family of Unimodal Maps Produce Finite Limit Cycles Monotonically with the Parameter?

In this note we consider a collection $\cal{C}$ of one parameter families of unimodal maps of $[0,1].$ Each family in the collection has the form $\{μf\}$ where $μ\in [0,1].$ Denoting the kneading sequence of $μf$ by $K(μf)$, we will prove that for each member of $\cal{C}$, the map $μ\mapsto K(μf)$ is monotone. It then follows that for each member of $\cal{C}$ the map $μ\mapsto h(μf)$ is monotone, where $h(uf)$ is the topological entropy of $μf.$ For interest, $μf(x)=4μx(1-x)$ and $μf(x)=μ\sin(πx)$ are shown to belong to $\cal{C}.$ This extends the work of Masato Tsujii [1].