Bayesian inference and learning in Gaussian process state-space models with Particle MCMC

State-space models are successfully used in many areas of science, engineering and economics to model time series and dynamical systems. We present a fully Bayesian approach to inference and learning (i.e. state estimation and system identification) in nonlinear nonparametric state-space models. We place a Gaussian process prior over the state transition dynamics, resulting in a flexible model able to capture complex dynamical phenomena. To enable efficient inference, we marginalize over the transition dynamics function and, instead, infer directly the joint smoothing distribution using specially tailored Particle Markov Chain Monte Carlo samplers. Once a sample from the smoothing distribution is computed, the state transition predictive distribution can be formulated analytically. Our approach preserves the full nonparametric expressivity of the model and can make use of sparse Gaussian processes to greatly reduce computational complexity.

Stress-induced domain dynamics and phase transitions in epitaxially grown VO₂ nanowires.

We demonstrate that surface stresses in epitaxially grown VO₂ nanowires (NWs) have a strong effect on the appearance and stability of intermediate insulating M₂ phases, as well as the spatial distribution of insulating and metallic domains during structural phase transitions. During the transition from an insulating M1 phase to a metallic R phase, the coexistence of insulating M₁ and M₂ phases with the absence of a metallic R phase was observed at atmospheric pressure. In addition, we show that, for a VO₂ NW without the presence of an epitaxial interface, surface stresses dominantly lead to spatially inhomogeneous phase transitions between insulating and metallic phases. In contrast, for a VO₂ NW with the presence of an epitaxial interface, the strong epitaxial interface interaction leads to additional stresses resulting in uniformly alternating insulating and metallic domains along the NW length.

Forecasting high-frequency futures returns using online langevin dynamics

Forecasting the returns of assets at high frequency is the key challenge for high-frequency algorithmic trading strategies. In this paper, we propose a jump-diffusion model for asset price movements that models price and its trend and allows a momentum strategy to be developed. Conditional on jump times, we derive closed-form transition densities for this model. We show how this allows us to extract a trend from high-frequency finance data by using a Rao-Blackwellized variable rate particle filter to filter incoming price data. Our results show that even in the presence of transaction costs our algorithm can achieve a Sharpe ratio above 1 when applied across a portfolio of 75 futures contracts at high frequency. © 2011 IEEE.

Twisting transition between crystalline and fibrillar phases of aggregated peptides

We study two distinctly ordered condensed phases of polypeptide molecules, amyloid fibrils and amyloidlike microcrystals, and the first-order twisting phase transition between these two states. We derive a single free-energy form which connects both phases. Our model identifies relevant degrees of freedom for describing the collective behavior of supramolecular polypeptide structures, reproduces accurately the results from molecular dynamics simulations as well as from experiments, and sheds light on the uniform nature of the dimensions of different peptide fibrils. © 2012 American Physical Society.

Aromaticity on the fly: cyclic transition state stabilization at finite temperature.

We study the transition state of pericyclic reactions at elevated temperature with unbiased ab initio molecular dynamics. We find that the transition state of the intramolecular rearrangements for barbaralane and bullvalene remains aromatic at high temperature despite the significant thermal atomic motions. Structural, magnetic, and electronic properties of the dynamical transition state show the concertedness and aromatic character. Free-energy calculations also support the validity of the transition state theory for the present rearrangement reactions. The calculations demonstrate that cyclic delocalization represents a strong force to synchronize the thermal atomic motions even at high temperatures.

Ultrafast dynamics of exciton formation in semiconductor nanowires.

The dynamics of free electron-hole pairs and excitons in GaAs-AlGaAs-GaAs core-shell-skin nanowires is investigated using femtosecond transient photoluminescence spectroscopy at 10 K. Following nonresonant excitation, a bimolecular interconversion of the initially generated electron-hole plasma into an exciton population is observed. This conducting-to-insulating transition appears to occur gradually over electron-hole charge pair densities of 2-4 × 10(16) cm(-3) . The smoothness of the Mott transition is attributed to the slow carrier-cooling during the bimolecular interconversion of free charge carriers into excitons and to the presence of chemical-potential fluctuations leading to inhomogeneous spectral characteristics. These results demonstrate that high-quality nanowires are model systems for investigating fundamental scientific effects in 1D heterostructures.