arXiv subjects

David Johnson

Publications and source records attributed to David Johnson.

25 records · Page 2Linked to original sources

arXiv2019

"Why did you do that?": Explaining black box models with Inductive Synthesis

By their nature, the composition of black box models is opaque. This makes the ability to generate explanations for the response to stimuli challenging. The importance of explaining black box models has become increasingly important given the prevalence of AI and ML systems and the need to build legal and regulatory frameworks around them. Such explanations can also increase trust in these uncertain systems. In our paper we present RICE, a method for generating explanations of the behaviour of black box models by (1) probing a model to extract model output examples using sensitivity analysis; (2) applying CNPInduce, a method for inductive logic program synthesis, to generate logic programs based on critical input-output pairs; and (3) interpreting the target program as a human-readable explanation. We demonstrate the application of our method by generating explanations of an artificial neural network trained to follow simple traffic rules in a hypothetical self-driving car simulation. We conclude with a discussion on the scalability and usability of our approach and its potential applications to explanation-critical scenarios.

cs.AI↗

arXiv2018

High efficiency laser-assisted H- charge exchange for microsecond duration beams

Laser-assisted stripping is a novel approach to H- charge exchange that overcomes long-standing limitations associated with the traditional, foil-based method of producing high-intensity, time-structured beams of protons. This paper reports on the first successful demonstration of the laser stripping technique for microsecond duration beams. The experiment represents a factor of 1000 increase in the stripped pulse duration compared with the previous proof-of-principle demonstration. The central theme of the experiment is the implementation of methods to reduce the required average laser power such that high efficiency stripping can be accomplished for microsecond duration beams using conventional laser technology. The experiment was performed on the Spallation Neutron Source 1 GeV H- beam using a 1 MW peak power UV laser and resulted in ~95% stripping efficiency.

physics.acc-ph↗

arXiv2016

Open and Regionalised Spectrum Repositories for Emerging Countries

TV White Spaces have recently been proposed as an alternative to alleviate the spectrum crunch, characterised by the need to reallocate frequency bands to accommodate the ever-growing demand for wireless communications. In this paper, we discuss the motivations and challenges for collecting spectrum measurements in developing regions and discuss a scalable system for communities to gather and provide access to White Spaces information through open and regionalised repositories. We further discuss two relevant aspects. First, we propose a cooperative mechanism for sensing spectrum availability using a detector approach. Second, we propose a strategy (and an architecture) on the database side to implement spectrum governance. Other aspects of the work include discussion of an extensive measurement campaign showing a number of white spaces in developing regions, an overview of our experience on low-cost spectrum analysers, and the architecture of zebra-rfo, an application for processing crowd-sourced spectrum data.

cs.NI↗

arXiv2015

Absence of Observed Unspeakably Large Black Holes Tells Us the Curvature of Space

Using black-hole arguments with widely accepted premises, we show that it is extremely improbable that space is Euclidean, and that it is unspeakably improbable that space is hyperbolic. Independently, using an argument which makes no appeal to black holes, but only to the ratio of volumes in hyperbolic space of Hubble volumes from different times, we prove that space is not hyperbolic. We conclude by discussing some implications of these results, when conjoined with the assumption that the entropy of a black hole is the hidden information about its internal configuration.

physics.hist-ph↗

arXiv2015

Design and development activities for in-vessel and in-port components of ITER microwave diagnostics

The ITER tokamak will be operating with 5 microwave diagnostic systems. While they rely on different physics, they share a common need: transmitting low and high power microwave in the range of 12 GHz to 1000 GHz(different bandwidths for different diagnostics) between the plasma and a diagnostic area tens of meters away. The designs proposed for vacuum windows, in-vessel waveguides and antennas are presented together with the development activities needed to finalise this work.

physics.plasm-ph↗

arXiv2014

Active Clustering with Model-Based Uncertainty Reduction

Semi-supervised clustering seeks to augment traditional clustering methods by incorporating side information provided via human expertise in order to increase the semantic meaningfulness of the resulting clusters. However, most current methods are \emph{passive} in the sense that the side information is provided beforehand and selected randomly. This may require a large number of constraints, some of which could be redundant, unnecessary, or even detrimental to the clustering results. Thus in order to scale such semi-supervised algorithms to larger problems it is desirable to pursue an \emph{active} clustering method---i.e. an algorithm that maximizes the effectiveness of the available human labor by only requesting human input where it will have the greatest impact. Here, we propose a novel online framework for active semi-supervised spectral clustering that selects pairwise constraints as clustering proceeds, based on the principle of uncertainty reduction. Using a first-order Taylor expansion, we decompose the expected uncertainty reduction problem into a gradient and a step-scale, computed via an application of matrix perturbation theory and cluster-assignment entropy, respectively. The resulting model is used to estimate the uncertainty reduction potential of each sample in the dataset. We then present the human user with pairwise queries with respect to only the best candidate sample. We evaluate our method using three different image datasets (faces, leaves and dogs), a set of common UCI machine learning datasets and a gene dataset. The results validate our decomposition formulation and show that our method is consistently superior to existing state-of-the-art techniques, as well as being robust to noise and to unknown numbers of clusters.

cs.LG↗

arXiv2012

Random Forests for Metric Learning with Implicit Pairwise Position Dependence

Metric learning makes it plausible to learn distances for complex distributions of data from labeled data. However, to date, most metric learning methods are based on a single Mahalanobis metric, which cannot handle heterogeneous data well. Those that learn multiple metrics throughout the space have demonstrated superior accuracy, but at the cost of computational efficiency. Here, we take a new angle to the metric learning problem and learn a single metric that is able to implicitly adapt its distance function throughout the feature space. This metric adaptation is accomplished by using a random forest-based classifier to underpin the distance function and incorporate both absolute pairwise position and standard relative position into the representation. We have implemented and tested our method against state of the art global and multi-metric methods on a variety of data sets. Overall, the proposed method outperforms both types of methods in terms of accuracy (consistently ranked first) and is an order of magnitude faster than state of the art multi-metric methods (16x faster in the worst case).

stat.ML↗