A Maximum-Likelihood Interpretation for Slow Feature Analysis

The brain extracts useful features from a maelstrom of sensory information, and a fundamental goal of theoretical neuroscience is to work out how it does so. One proposed feature extraction strategy is motivated by the observation that the meaning of sensory data, such as the identity of a moving visual object, is often more persistent than the activation of any single sensory receptor. This notion is embodied in the slow feature analysis (SFA) algorithm, which uses “slowness” as an heuristic by which to extract semantic information from multi-dimensional time-series. Here, we develop a probabilistic interpretation of this algorithm showing that inference and learning in the limiting case of a suitable probabilistic model yield exactly the results of SFA. Similar equivalences have proved useful in interpreting and extending comparable algorithms such as independent component analysis. For SFA, we use the equivalent probabilistic model as a conceptual spring-board, with which to motivate several novel extensions to the algorithm.

Shock / boundary-layer interaction control using three-dimensional bumps for transonic wings

Three-dimensional bumps have been developed and investigated, aiming at the two major objectives of shock-wave / boundary-layer interaction control, i.e. drag reduction and suppression of separation, simultaneously. An experimental investigation has been conducted for a default rounded bump in channel now at University of Cambridge and a computational study has been performed for a spanwise series of rounded bumps mounted on a transonic aerofoil at University of Stuttgart. Observed in both cases are wave drag reduction owing to A-shock structures produced by three-dimensional surface bumps and mild control effects on the boundary layer. The effects of rough surface and tall extension have been investigated as well as several geometric variations and multiple bump configurations. A double configuration of narrow rounded bumps has been found to best perform amongst the tested, considerably reducing wave drag through a well-established A-shock structure with little viscous penalty and thus achieving substantial overall drag reduction. Counter-rotating streamwise vortex pairs have been produced by some configurations as a result of local flow separation, but they have been observed to be confined in relatively narrow wake regions, expected to be beneficial in suppressing large-scale separation under off-design condition despite increase of viscous drag. On the whole a large potential of three-dimensional control with discrete rounded bumps has been demonstrated both experimentally and numerically, and experimental investigation of bumps fitted on a transonic aerofoil or wing is suggested toward practical application.

Transient thermal analysis of power devices based on Fourier-series thermal model

A new thermal model based on Fourier series expansion method has been presented for dynamic thermal analysis on power devices. The thermal model based on the Fourier series method has been programmed in MATLAB SIMULINK and integrated with a physics-based electrical model previously reported. The model was verified for accuracy using a two-dimensional Fourier model and a two-dimensional finite difference model for comparison. To validate this thermal model, experiments using a 600V 50A IGBT module switching an inductive load, has been completed under high frequency operation. The result of the thermal measurement shows an excellent match with the simulated temperature variations and temperature time-response within the power module. ©2008 IEEE.

Multilinear function factorisation for time series feature extraction

This work applies a variety of multilinear function factorisation techniques to extract appropriate features or attributes from high dimensional multivariate time series for classification. Recently, a great deal of work has centred around designing time series classifiers using more and more complex feature extraction and machine learning schemes. This paper argues that complex learners and domain specific feature extraction schemes of this type are not necessarily needed for time series classification, as excellent classification results can be obtained by simply applying a number of existing matrix factorisation or linear projection techniques, which are simple and computationally inexpensive. We highlight this using a geometric separability measure and classification accuracies obtained though experiments on four different high dimensional multivariate time series datasets. © 2013 IEEE.