Improving neural network training solutions using regularisation

This paper describes the application of regularisation to the training of feedforward neural networks, as a means of improving the quality of solutions obtained. The basic principles of regularisation theory are outlined for both linear and nonlinear training and then extended to cover a new hybrid training algorithm for feedforward neural networks recently proposed by the authors. The concept of functional regularisation is also introduced and discussed in relation to MLP and RBF networks. The tendency for the hybrid training algorithm and many linear optimisation strategies to generate large magnitude weight solutions when applied to ill-conditioned neural paradigms is illustrated graphically and reasoned analytically. While such weight solutions do not generally result in poor fits, it is argued that they could be subject to numerical instability and are therefore undesirable. Using an illustrative example it is shown that, as well as being beneficial from a generalisation perspective, regularisation also provides a means for controlling the magnitude of solutions. (C) 2001 Elsevier Science B.V. All rights reserved.

Improving f(MAX)/f(T) ratio in FinFETs using source/drain extension region engineering

A design methodology to optimise the ratio of maximum oscillation frequency to cutoff frequency, f(MAX)/f(T), in 60 nm FinFETs is presented. Results show that 25 to 60% improvement in f(MAX)/f(T) at drain currents of 20-300 mu A/mu m can be achieved in a non-overlap gate-source/drain architecture. The reported work provides new insights into the design and optimisation of nanoscale FinFETs for RF applications.

Lean principles and techniques for improving the quality and productivity of software development projects: a case study

Software product development is recognised as difficult due to the intangible nature of the product, requirements elicitation, effective progress measurement, and so forth. In this paper, we describe some of the challenges of software product development and how the challenges are being met by lean management principles and techniques. Specifically, we examine lean principles and techniques that were devised by Toyota and other manufacturers over the last 50 years. Applying lean principles to software development projects has been advocated for over ten years and it will be shown that the extensive lean literature is a valuable source of ideas for software development. A case study with a software development organisation, Timberline Inc., will demonstrate that lean principles and techniques can be successfully applied to software product development.

Choice and use of standards for dynamic headspace trapping and application to the analysis of the volatiles of baked potato

When attempting to quantify the volatile components of a food isolated by dynamic headspace trapping onto an adsorbent, the analyst has to select the most appropriate compounds to use as standards and at which stage of the analysis to add them. Factors to be borne in mind include the volatility of the standard, the response of the GC detector, and whether to add the standard to the sample or to the adsorbent trap. This chapter considers the issues and describes the application of one chosen method to the quantitation of the volatile components of baked potato.

Improving the utilisation of University teaching space

There is a perception that teaching space in universities is a rather scarce resource. However, some studies have revealed that in many institutions it is actually chronically under-used. Often, rooms are occupied only half the time, and even when in use they are often only half full. This is usually measured by the “utilisation” which is basically the percentage of available ’seat-hours’ that are employed. In real institutions, this utilisation can often takes values as low as 20-40%. One consequence of such low utilisation is that space managers are under pressure to make a more efficient use of the available teaching space. However, better management is hampered because there does not appear to be a good understanding within space management (near-term planning) of why this happens. Nor, a good basis within space planning (long-term planning) of how best to accommodate the expected low utilisations. This motivates our two main goals: (i) To understand the factors that drive down utilisations, (ii) To set up methods to provide better space planning. Here, we provide quantitative evidence that constraints arising from timetabling and location requirements easily have the potential to explain the low utilisations seen in reality. Furthermore, on considering the decision question “Can this given set of courses all be allocated in the available teaching space?” we find that the answer depends on the associated utilisation in a way that exhibits threshold behaviour: There is a sharp division between regions in which the answer is “almost always yes” and those of “almost always no”. Our work suggests that progress in space management and planning will arise from an integrated approach; combining purely space issues with restrictions representing an aggregated or abstracted version of key constraints such as timetabling or location, and

Casting of Self-Compacting Concrete

Increased productivity and improved working environment have had high priority in the development of concrete construction over the last decade. Development of a material not needing vibration for compaction—i.e. selfcompacting concrete (SCC)—has successfully met the challenge and is now increasingly being used in routine practice. The key to the improvement of fresh concrete performance has been nanoscale tailoring of molecules for surface active admixtures, as well as improved understanding of particle packing and of the role of mineral surfaces in cementitious matrixes. Fundamental studies of rheological behaviour of cementitious particle suspensions were soon expanded to extensive innovation programmes incorporating applied research, site experiments, instrumented full scale applications supporting technology, standards and guides, information efforts as well as training programmes. The major impact of the introduction of SCC is connected to the production process. The choice and handling of constituents are modified as well as mix design, batching, mixing and transporting. The productivity is drastically improved through elimination of vibration compaction and process reorganisation. The working environment is significantly enhanced through avoidance of vibration induced damages, reduced noise and improved safety. Additionally, the technology is improving performance in terms of hardened material properties like surface quality, strength and durability.