arXiv subjects

Terry Bossomaier

Publications and source records attributed to Terry Bossomaier.

8 recordsLinked to original sources

arXiv2021

Information Flow in First-Order Potts Model Phase Transition

Phase transitions abound in nature and society, and, from species extinction to stock market collapse, their prediction is of widespread importance. In earlier work we showed that Global Transfer Entropy, a general measure of information flow, was found to peak away from the transition on the disordered side for the Ising model, a canonical second-order transition. Here we show that (a) global transfer entropy also peaks on the disordered side of the transition of finite first-order transitions, i.e., those which have finite latent heat and no correlation length divergence, such as ecology dynamics on coral reefs, and (b) analysis of information flow across state boundaries unifies both transition orders. We obtain the first information-theoretic result for the high-order Potts model and the first demonstration of early warning of a first-order transition. The unexpected earlier finding that global transfer entropy peaks on the disordered side of a transition is also found for finite first-order systems, albeit not in the thermodynamic limit. By noting that the interface length of clusters in each phase is the dominant region of information flow, we unify the information theoretic behaviour of first and second-order transitions.

cond-mat.stat-mech↗

arXiv2018

Information Flow in Finite Flocks

We simulate the canonical Vicsek model and estimate the flow of information as a function of noise (the variability in the extent to which each animal aligns with its neighbours). We show that the global transfer entropy for finite flocks not only fails to peak near the phase transition, as demonstrated for the canonical 2D Ising model, but remains constant from the transition to very low noise values. This provides a foundation for future study regarding information flow in more complex models and real-world flocking data.

cond-mat.stat-mech↗

arXiv2018

Information Flow in Finite Flocks with Topological Interactions

We simulate the Vicsek model utilising topological neighbour interactions and estimate information theoretic quantities as a function of noise, the variability in the extent to which each animal aligns with its neighbours, and the flock direction. We show that these quantities, mutual information and global transfer entropy, are in fact dependent on observation time, and in comparison to the canonical Vicsek model which utilises range-based interactions, the topological variant converges to the long-term limiting behaviour with smaller observation windows. Finally, we show that in contrast to the metric model, which exhibits maximal information flow for the ordered regime, the topological model maintains this maximal information flow beyond the phase transition and into the disordered regime.

cond-mat.stat-mech↗

arXiv2018

Did William Shakespeare and Thomas Kyd Write Edward III?

William Shakespeare is believed to be a significant author in the anonymous play, The Reign of King Edward III, published in 1596. However, recently, Thomas Kyd, has been suggested as the primary author. Using a neurolinguistics approach to authorship identification we use a four-feature technique, RPAS, to convert the 19 scenes in Edward III into a multi-dimensional vector. Three complementary analytical techniques are applied to cluster the data and reduce single technique bias before an alternate method, seriation, is used to measure the distances between clusters and test the strength of the connections. We find the multivariate techniques robust and are able to allocate up to 14 scenes to Thomas Kyd, and further question if scenes long believed to be Shakespeare's are not his.

cs.CL↗

arXiv2017

Review of Data Structures for Computationally Efficient Nearest-Neighbour Entropy Estimators for Large Systems with Periodic Boundary Conditions

Information theoretic quantities are extremely useful in discovering relationships between two or more data sets. One popular method---particularly for continuous systems---for estimating these quantities is the nearest neighbour estimators. When system sizes are very large or the systems have periodic boundary conditions issues with performance and correctness surface, however solutions are known for each problem. Here we show that these solutions are inappropriate in systems that simultaneously contain both features and discuss a lesser known alternative solution involving Vantage Point trees that is capable of addressing both issues.

stat.CO↗

arXiv2017

Anomalous Behaviour of Mutual Information in Finite Flocks

The existing consensus is that flocks are poised at criticality, entailing long correlation lengths and a maximal value of Shannon mutual information in the large-system limit. We show, by contrast, that for finite flocks in the long observation time limit, mutual information may not only fail to peak at criticality---as observed for other critical systems---but also diverge as noise tends to zero. This result carries implications for other finite-size, out-of-equilibrium systems, where observation times may vary widely compared to time scales of internal system dynamics; thus it may not be assumed that mutual information locates the phase transition.

cond-mat.stat-mech↗

arXiv2012

Transfer Entropy as a Log-likelihood Ratio

Transfer entropy, an information-theoretic measure of time-directed information transfer between joint processes, has steadily gained popularity in the analysis of complex stochastic dynamics in diverse fields, including the neurosciences, ecology, climatology and econometrics. We show that for a broad class of predictive models, the log-likelihood ratio test statistic for the null hypothesis of zero transfer entropy is a consistent estimator for the transfer entropy itself. For finite Markov chains, furthermore, no explicit model is required. In the general case, an asymptotic chi-squared distribution is established for the transfer entropy estimator. The result generalises the equivalence in the Gaussian case of transfer entropy and Granger causality, a statistical notion of causal influence based on prediction via vector autoregression, and establishes a fundamental connection between directed information transfer and causality in the Wiener-Granger sense.

stat.AP↗

arXiv1998

Evolution of Neural Networks to Play the Game of Dots-and-Boxes

Dots-and-Boxes is a child's game which remains analytically unsolved. We implement and evolve artificial neural networks to play this game, evaluating them against simple heuristic players. Our networks do not evaluate or predict the final outcome of the game, but rather recommend moves at each stage. Superior generalisation of play by co-evolved populations is found, and a comparison made with networks trained by back-propagation using simple heuristics as an oracle.

cs.NE↗