Online learning via congregational gradient descent

We propose and analyse a populational version of stepwise gradient descent suitable for a wide range of learning problems. The algorithm is motivated by genetic algorithms which update a population of solutions rather than just a single representative as is typical for gradient descent. This modification of traditional gradient descent (as used, for example, in the backpropogation algorithm) avoids getting trapped in local minima. We use an averaging analysis of the algorithm to relate its behaviour to an associated ordinary differential equation. We derive a result concerning how long one has to wait in order that, with a given high probability, the algorithm is within a certain neighbourhood of the global minimum. We also analyse the effect of different population sizes. An example is presented which corroborates our theory very well.This work was supported by the Australian Research Council.     

Nonlinear averaging theorems, and the determination of parameter convergence rates in adaptive control

The paper presents nonlinear averaging theorems for two-time scale systems, where the dynamics of the fast system are allowed to vary with the slow system. The results are applied to the Narendra-Valavani adaptive control algorithm, and estimates of the parameter convergence rates are obtained which do not rely on a linearization of the system around the equilibrium, and therefore are valid in a larger region in the parameter space.     

Comparability problems in the analysis of multiway data

Almost all methods for the analysis of multiway data assume that the comparison of any two entries in the data array under study reflects or represents meaningful content-specific information. This is especially the case if one wants the data analysis to yield insight into the real mechanisms underlying the data. Violation of this assumption may imply data-analytic results that are of doubtful quality at best and worthless in the worst-case scenario. In the present paper, we first clarify why comparability is a key assumption in most methods for multiway data analysis. Next, we list a number of reasons why this assumption is very often violated in practice. We then review a few possible approaches that have been advanced to deal with problems of comparability, and discuss their advantages and shortcomings. We conclude by clarifying that any satisfactory solution to comparability problems requires a very careful reflection about the data collection and the ultimate goal of the data analysis.     

A unifying model involving a categorical and/or dimensional reduction for multimode data

A unifying model is presented that implies a categorical and/or dimensional reduction of one or several modes of a multiway data set. The model encompasses a broad range of (existing as well as to be developed) discrete, continuous, as well as hybrid discrete-continuous reduction models as special cases, which all imply a decomposition of the reconstructed data on the basis of quantifications of the different data modes and a linking array. An analysis of the objective or loss function associated with the model leads to two generic algorithmic strategies, the possibilities and limitations of which are the object of a subsequent discussion.     

Individual differences in core affect variability and their relationship to personality and psychological adjustment.

How people's feelings change across time can be represented as trajectories in a core affect space defined by the dimensions of valence and activation. In this article, the authors analyzed individual differences in within-person affective variability defined as characteristics of core affect trajectories, introducing new ways to conceptualize affective variability. In 2 studies, participants provided multiple reports across time describing how they were feeling in terms of core affect. From these data, characteristics of participants' core affect trajectories were derived. Across both studies, core affect variability was negatively related to average valence, self-esteem, and agreeableness, and it was positively related to neuroticism and depression. Moreover, spin, a measure of how much people experienced qualitatively different feelings within the core affect space, was related more consistently to trait measures of adjustment and personality than other measures of within-person variability, including widely used measures of within-person single-dimension standard deviations. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Open-loop discretization methods for control systems design

In this paper, a unified theory is presented which addresses essential aspects of the open-loop discretization procedure. Initially, factors which affect the generation and propagation of discretization errors are identified by analytical, heuristic and experimental arguments. Following this, a discretization algorithm is presented which takes these factors into account. The fundamental idea of the discretization method is the replacement of the analogue integrators of the prototype continuous-time system by discrete-time approximations. This is done in such a way as to optimize a given cost function with respect to a given input. Unlike many discretization schemes presented in the literature, the procedure developed attempts to first determine the magnitude of the expected discretization errors, and then discretize with an appropriate complexity, giving the designer control over the order of the discrete-time system. This appears to be an effective means of designing filters with low complexity which still retain essential properties of the prototype system. An emphasis of this work is to view the discretization process from a control theory and engineering perspective. This results in a number of new perspectives relating to the discretization process.     

The appraisal basis of anger: Specificity, necessity and sufficiency of components.

The nature of the association between anger and 5 appraisal-action tendency components--goal obstacle, other accountability, unfairness, control, and antagonism--was examined in terms of specificity, necessity, and sufficiency. In 2 studies, participants described recently experienced unpleasant situations in which 1 of the appraisal-action tendency components was present or absent and indicated which emotions they had experienced. The results showed that (a) other accountability and arrogant entitlement, as an instance of unfairness, are specific appraisals ability for anger; and most important, (b) none of the components is necessary or sufficient for anger. The findings suggest that the relation between emotions and appraisal-action tendency components should be conceptualized instead as a contingent association, meaning that they usually co-occur. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A new optimal design and analysis method based on MADM for MEMS products development

This paper presents an optimum design method to support the total micro-electromechanical systems (MEMS) product/device optimization, and its evaluation at the conceptual stage itself using the multiple attribute decision making method. In the traditional MEMS product development cycle, simulation and design using software tools are very important due to the knowledge limitation and complexity in design, fabrication, and packaging processes. The available tools are time consuming and relay on trial and error to achieve an optimum solution. The proposed method simplifies the relationship between parameters of design, fabrication, materials, packaging, and the performance of the MEMS product. The methodology is explained with the help of design flow diagram and time chart. A MEMS-based radio frequency (RF) power sensor is designed and the methodology is demonstrated. The proposed sensitivity analysis method is more effective and less time consuming than traditional techniques. Sensitivity analysis is carried out by varying the thickness of the signal conductor. The results of RF power sensor with insertion loss 0.428?dB, reflection loss 25.956 and voltage standing wave ratio of 1.106 at 1.5?GHz are reported.