Adaptive back-propagation in on-line learning of multilayer networks

An adaptive back-propagation algorithm is studied and compared with gradient descent (standard back-propagation) for on-line learning in two-layer neural networks with an arbitrary number of hidden units. Within a statistical mechanics framework, both numerical studies and a rigorous analysis show that the adaptive back-propagation method results in faster training by breaking the symmetry between hidden units more efficiently and by providing faster convergence to optimal generalization than gradient descent.

Finite-size effects in on-line learning of multilayer neural networks

We complement recent advances in thermodynamic limit analyses of mean on-line gradient descent learning dynamics in multi-layer networks by calculating fluctuations possessed by finite dimensional systems. Fluctuations from the mean dynamics are largest at the onset of specialisation as student hidden unit weight vectors begin to imitate specific teacher vectors, increasing with the degree of symmetry of the initial conditions. In light of this, we include a term to stimulate asymmetry in the learning process, which typically also leads to a significant decrease in training time.

Die Poetik des Extremen. Ausschreitungen einer Sprache des Radikalen:excesses of a language of radical

Gibt es eine Traditionslinie extremistischer Poetiken in der deutschsprachigen Literatur? Uwe Schütte untersucht anhand literarischer Texte ab dem späten 18. Jahrhundert den Konnex zwischen historischen Phänomenen wie Revolution, Krieg oder Terrorismus und extremen biografischen Umständen wie Schizophrenie für die Herausbildung radikaler Schreibweisen. Die Spannbreite der behandelten Autoren reicht dabei von der Klassikertrias Kleist, Hölderlin und Büchner über Schriftsteller des 20. Jahrhunderts wie Ernst Jünger oder Hans Henny Jahnn bis zu den Gegenwartsautoren Ernst Herbeck und Rainald Goetz. The study investigates aesthetic representations of extremism in German-language literature from around 1800 to the present. Its aim is to examine the interplay between three different areas: historical circumstances, (auto)biographical issues, and literary texts. Discussed are texts by both major and marginal writers from various genres, ranging from classics such as Heinrich von Kleist or Friedrich Hölderlin to the marginalised poet Ernst Herbeck or the contemporary writer Rainald Goetz. Subjects and factors considered include extremist phenomena in modern history (such as revolutions, wars, terrorism) and extreme individual experiences (such as suicide or schizophrenia) on the aesthetic domain(s) with regard to the production of literary discourses that could be considered as extremist. These manifest themselves in the development of what can be viewed as ‘radical poetics’, decidedly innovative styles of writing and moral or political transgression in fiction. Being the first critical attempt to trace the history of radical discourses in German literature, the study explores the validity of creating an aesthetic category of 'literary extremism'.

The role of biases in on-line learning of two-layer networks

The influence of biases on the learning dynamics of a two-layer neural network, a normalized soft-committee machine, is studied for on-line gradient descent learning. Within a statistical mechanics framework, numerical studies show that the inclusion of adjustable biases dramatically alters the learning dynamics found previously. The symmetric phase which has often been predominant in the original model all but disappears for a non-degenerate bias task. The extended model furthermore exhibits a much richer dynamical behavior, e.g. attractive suboptimal symmetric phases even for realizable cases and noiseless data.

On-line learning of unrealizable tasks

The dynamics of on-line learning is investigated for structurally unrealizable tasks in the context of two-layer neural networks with an arbitrary number of hidden neurons. Within a statistical mechanics framework, a closed set of differential equations describing the learning dynamics can be derived, for the general case of unrealizable isotropic tasks. In the asymptotic regime one can solve the dynamics analytically in the limit of large number of hidden neurons, providing an analytical expression for the residual generalization error, the optimal and critical asymptotic training parameters, and the corresponding prefactor of the generalization error decay.

Semi-supervised learning of hierarchical latent trait models for data visualisation

An interactive hierarchical Generative Topographic Mapping (HGTM) ¸iteH_GTM has been developed to visualise complex data sets. In this paper, we build a more general visualisation system by extending the HGTM visualisation system in 3 directions: bf (1) We generalize HGTM to noise models from the exponential family of distributions. The basic building block is the Latent Trait Model (LTM) developed in ¸iteKaban_pami. bf (2) We give the user a choice of initializing the child plots of the current plot in either em interactive, or em automatic mode. In the interactive mode the user interactively selects ``regions of interest'' as in ¸iteH_GTM, whereas in the automatic mode an unsupervised minimum message length (MML)-driven construction of a mixture of LTMs is employed. bf (3) We derive general formulas for magnification factors in latent trait models. Magnification factors are a useful tool to improve our understanding of the visualisation plots, since they can highlight the boundaries between data clusters. The unsupervised construction is particularly useful when high-level plots are covered with dense clusters of highly overlapping data projections, making it difficult to use the interactive mode. Such a situation often arises when visualizing large data sets. We illustrate our approach on a toy example and apply our system to three more complex real data sets.

When in Rome ... ?:Human resource management and the performance of foreign firms operating in India

Purpose: The purpose of the paper is to examine the kind of HRM practices being implemented by overseas firms in their Indian subsidiaries and also to analyze the linkage between HRM practices and organizational performance. Design/methodology/approach: The paper utilizes a mixture of both quantitative and qualitative techniques via personal interviews in 76 subsidiaries. Findings: The results show that while the introduction of HRM practices from the foreign parent organization is negatively associated with performance, local adaption of HRM practices is positively related with the performance of foreign firms operating in India. Research limitations/implications: The main limitations include data being collected by only one respondent from each firm, and the relatively small sample size. Practical implications: The key message for practitioners is that HRM systems do improve organizational performance in the Indian subsidiaries of foreign firms, and an emphasis on the localization of HRM practices can further contribute in this regard. Originality/value: This is perhaps the very first investigation of its kind in the Indian context. © Emerald Group Publishing Limited.

The impact of foreign direct investment on UK manufacturing: is there a profit squeeze in domestic firms?

This paper examines the extent to which foreign direct investment (FDI) in selected UK manufacturing sectors has an impact on reported profits in domestic firms. Foreign manufacturing firms are characterized by relatively high labour productivity and low wage shares. Entry by foreign firms not only impacts on domestic market shares, but also on domestic cost conditions. As a result, profitability in the indigenous sector may be reduced. There are a number of policy implications of this analysis which are explored.

Indirect employment effects of Foreign Direct Investment into the UK

This paper examines some of the employment consequences, broadly defined, associated with foreign inward investment. A foreign firm entering an industry in the UK will have a degree of firm-specific advantage oover the incumbent firms. This advantage is assumed to manifest itself in terms of a productivity differential over the domestic sector. As such, foreign entry will create factor market disequlibrium in the domestic sector. It is shown that such investment generates 'employment substitution' away from UK firms, equivalent to approximately one-fifth of all the jobs created by inward investment.

The effects of foreign acquisition on domestic and export markets dynamics in China

Using recent data from the Chinese manufacturing industry and the generalised propensity score, this paper establishes economically significant causal effects of foreign acquisition on domestic and export markets dynamics.

What determines innovation activity in Chinese state-owned enterprises? The role of foreign direct investment

We investigate whether inward foreign direct investment (FDI), either at the firm or industry level, has any impact on product innovation by Chinese state-owned enterprises (SOEs). We use a comprehensive firm-level panel data set of some 20,000 SOEs during 1999-2005. Our results show that foreign capital participation at the firm level is associated with higher innovative activity. Inward FDI in the sector, by contrast, has a negative effect on innovative activity in SOEs on average. However, there is a positive effect of sector-level FDI on SOEs that export, invest in human capital, or undertake R&D. © 2008 Elsevier Ltd. All rights reserved.