Regularization and complexity control in feed-forward networks

In this paper we consider four alternative approaches to complexity control in feed-forward networks based respectively on architecture selection, regularization, early stopping, and training with noise. We show that there are close similarities between these approaches and we argue that, for most practical applications, the technique of regularization should be the method of choice.

Blockade of interleukin-6 signaling inhibits the classic pathway and promotes an alternative pathway of macrophage activation after spinal cord injury in mice

Background Recent in vivo and in vitro studies in non-neuronal and neuronal tissues have shown that different pathways of macrophage activation result in cells with different properties. Interleukin (IL)-6 triggers the classically activated inflammatory macrophages (M1 phenotype), whereas the alternatively activated macrophages (M2 phenotype) are anti-inflammatory. The objective of this study was to clarify the effects of a temporal blockade of IL-6/IL-6 receptor (IL-6R) engagement, using an anti-mouse IL-6R monoclonal antibody (MR16-1), on macrophage activation and the inflammatory response in the acute phase after spinal cord injury (SCI) in mice. Methods MR16-1 antibodies versus isotype control antibodies or saline alone were administered immediately after thoracic SCI in mice. SC tissue repair was compared between the two groups by Luxol fast blue (LFB) staining for myelination and immunoreactivity for the neuronal markers growth-associated protein (GAP)-43 and neurofilament heavy 200 kDa (NF-H) and for locomotor function. The expression of T helper (Th)1 cytokines (interferon (IFN)-? and tumor necrosis factor-a) and Th2 cytokines (IL-4, IL-13) was determined by immunoblot analysis. The presence of M1 (inducible nitric oxide synthase (iNOS)-positive, CD16/32-positive) and M2 (arginase 1-positive, CD206-positive) macrophages was determined by immunohistology. Using flow cytometry, we also quantified IFN-? and IL-4 levels in neutrophils, microglia, and macrophages, and Mac-2 (macrophage antigen-2) and Mac-3 in M2 macrophages and microglia. Results LFB-positive spared myelin was increased in the MR16-1-treated group compared with the controls, and this increase correlated with enhanced positivity for GAP-43 or NF-H, and improved locomotor Basso Mouse Scale scores. Immunoblot analysis of the MR16-1-treated samples identified downregulation of Th1 and upregulation of Th2 cytokines. Whereas iNOS-positive, CD16/32-positive M1 macrophages were the predominant phenotype in the injured SC of non-treated control mice, MR16-1 treatment promoted arginase 1-positive, CD206-positive M2 macrophages, with preferential localization of these cells at the injury site. MR16-1 treatment suppressed the number of IFN-?-positive neutrophils, and increased the number of microglia present and their positivity for IL-4. Among the arginase 1-positive M2 macrophages, MR16-1 treatment increased positivity for Mac-2 and Mac-3, suggestive of increased phagocytic behavior. Conclusion The results suggest that temporal blockade of IL-6 signaling after SCI abrogates damaging inflammatory activity and promotes functional recovery by promoting the formation of alternatively activated M2 macrophages.

Interocular suppression and contrast gain control in human vision

The human visual system combines contrast information from the two eyes to produce a single cyclopean representation of the external world. This task requires both summation of congruent images and inhibition of incongruent images across the eyes. These processes were explored psychophysically using narrowband sinusoidal grating stimuli. Initial experiments focussed on binocular interactions within a single detecting mechanism, using contrast discrimination and contrast matching tasks. Consistent with previous findings, dichoptic presentation produced greater masking than monocular or binocular presentation. Four computational models were compared, two of which performed well on all data sets. Suppression between mechanisms was then investigated, using orthogonal and oblique stimuli. Two distinct suppressive pathways were identified, corresponding to monocular and dichoptic presentation. Both pathways impact prior to binocular summation of signals, and differ in their strengths, tuning, and response to adaptation, consistent with recent single-cell findings in cat. Strikingly, the magnitude of dichoptic masking was found to be spatiotemporally scale invariant, whereas monocular masking was dependent on stimulus speed. Interocular suppression was further explored using a novel manipulation, whereby stimuli were presented in dichoptic antiphase. Consistent with the predictions of a computational model, this produced weaker masking than in-phase presentation. This allowed the bandwidths of suppression to be measured without the complicating factor of additive combination of mask and test. Finally, contrast vision in strabismic amblyopia was investigated. Although amblyopes are generally believed to have impaired binocular vision, binocular summation was shown to be intact when stimuli were normalized for interocular sensitivity differences. An alternative account of amblyopia was developed, in which signals in the affected eye are subject to attenuation and additive noise prior to binocular combination.

Motion sharpening and contrast:Gain control precedes compressive non-linearity?

Blurred edges appear sharper in motion than when they are stationary. We (Vision Research 38 (1998) 2108) have previously shown how such distortions in perceived edge blur may be accounted for by a model which assumes that luminance contrast is encoded by a local contrast transducer whose response becomes progressively more compressive as speed increases. If the form of the transducer is fixed (independent of contrast) for a given speed, then a strong prediction of the model is that motion sharpening should increase with increasing contrast. We measured the sharpening of periodic patterns over a large range of contrasts, blur widths and speeds. The results indicate that whilst sharpening increases with speed it is practically invariant with contrast. The contrast invariance of motion sharpening is not explained by an early, static compressive non-linearity alone. However, several alternative explanations are also inconsistent with these results. We show that if a dynamic contrast gain control precedes the static non-linear transducer then motion sharpening, its speed dependence, and its invariance with contrast, can be predicted with reasonable accuracy. © 2003 Elsevier Science Ltd. All rights reserved.

When do firms conform? The effect of regulatory control processes on compliance and opportunism

The authors use social control theory to develop a conceptual model that addresses the effectiveness of regulatory agencies’ (e.g., Food and Drug Administration, Occupational Safety and Health Administration) field-level efforts to obtain conformance with product safety laws. Central to the model are the control processes agencies use when monitoring organizations and enforcing the safety rules. These approaches can be labeled formal control (e.g., rigid enforcement) and informal control (e.g., social instruction). The theoretical framework identifies an important antecedent of control and the relative effectiveness of control’s alternative forms in gaining compliance and reducing opportunism. Furthermore, the model predicts that the regulated firms’ level of agreement with the safety rules moderates the relationships between control and firm responses. A local health department’s administration of state food safety regulations provides the empirical context for testing the hypotheses. The results from a survey of 173 restaurants largely support the proposed model. The study findings inform a discussion of effective methods of administering product safety laws. The authors use social control theory to develop a conceptual model that addresses the effectiveness of regulatory agencies’ (e.g., Food and Drug Administration, Occupational Safety and Health Administration) field-level efforts to obtain conformance with product safety laws. Central to the model are the control processes agencies use when monitoring organizations and enforcing the safety rules. These approaches can be labeled formal control (e.g., rigid enforcement) and informal control (e.g., social instruction). The theoretical framework identifies an important antecedent of control and the relative effectiveness of control’s alternative forms in gaining compliance and reducing opportunism. Furthermore, the model predicts that the regulated firms’ level of agreement with the safety rules moderates the relationships between control and firm responses. A local health department’s administration of state food safety regulations provides the empirical context for testing the hypotheses. The results from a survey of 173 restaurants largely support the proposed model. The study findings inform a discussion of effective methods of administering product safety laws.

Adaptation and gain pool summation:Alternative models and masking data

Foley [J. Opt. Soc. Am. A 11 (1994) 1710] has proposed an influential psychophysical model of masking in which mask components in a contrast gain pool are raised to an exponent before summation and divisive inhibition. We tested this summation rule in experiments in which contrast detection thresholds were measured for a vertical 1 c/deg (or 2 c/deg) sine-wave component in the presence of a 3 c/deg (or 6 c/deg) mask that had either a single component oriented at -45° or a pair of components oriented at ±45°. Contrary to the predictions of Foley's model 3, we found that for masks of moderate contrast and above, threshold elevation was predicted by linear summation of the mask components in the inhibitory stage of the contrast gain pool. We built this feature into two new models, referred to as the early adaptation model and the hybrid model. In the early adaptation model, contrast adaptation controls a threshold-like nonlinearity on the output of otherwise linear pathways that provide the excitatory and inhibitory inputs to a gain control stage. The hybrid model involves nonlinear and nonadaptable routes to excitatory and inhibitory stages as well as an adaptable linear route. With only six free parameters, both models provide excellent fits to the masking and adaptation data of Foley and Chen [Vision Res. 37 (1997) 2779] but unlike Foley and Chen's model, are able to do so with only one adaptation parameter. However, only the hybrid model is able to capture the features of Foley's (1994) pedestal plus orthogonal fixed mask data. We conclude that (1) linear summation of inhibitory components is a feature of contrast masking, and (2) that the main aftereffect of spatial adaptation on contrast increment thresholds can be assigned to a single site. © 2002 Elsevier Science Ltd. All rights reserved.

Computer integrated manufacturing control

Many manufacturing companies have long endured the problems associated with the presence of `islands of automation'. Due to rapid computerisation, `islands' such as Computer-Aided Design (CAD), Computer-Aided Manufacturing (CAM), Flexible Manufacturing Systems (FMS) and Material Requirement Planning (MRP), have emerged, and with a lack of co-ordination, often lead to inefficient performance of the overall system. The main objective of Computer-Integrated Manufacturing (CIM) technology is to form a cohesive network between these islands. Unfortunately, a commonly used approach - the centralised system approach, has imposed major technical constraints and design complication on development strategies. As a consequence, small companies have experienced difficulties in participating in CIM technology. The research described in this thesis has aimed to examine alternative approaches to CIM system design. Through research and experimentation, the cellular system approach, which has existed in the form of manufacturing layouts, has been found to simplify the complexity of an integrated manufacturing system, leading to better control and far higher system flexibility. Based on the cellular principle, some central management functions have also been distributed to smaller cells within the system. This concept is known, specifically, as distributed planning and control. Through the development of an embryo cellular CIM system, the influence of both the cellular principle and the distribution methodology have been evaluated. Based on the evidence obtained, it has been concluded that distributed planning and control methodology can greatly enhance cellular features within an integrated system. Both the cellular system approach and the distributed control concept will therefore make significant contributions to the design of future CIM systems, particularly systems designed with respect to small company requirements.

On waveform analysis, the adaptive task technique, and the effect of vibration on manual control skill

A re-examination of fundamental concepts and a formal structuring of the waveform analysis problem is presented in Part I. eg. the nature of frequency is examined and a novel alternative to the classical methods of detection proposed and implemented which has the advantage of speed and independence from amplitude. Waveform analysis provides the link between Parts I and II. Part II is devoted to Human Factors and the Adaptive Task Technique. The Historical, Technical and Intellectual development of the technique is traced in a review which examines the evidence of its advantages relative to non-adaptive fixed task methods of training, skill assessment and man-machine optimisation. A second review examines research evidence on the effect of vibration on manual control ability. Findings are presented in terms of percentage increment or decrement in performance relative to performance without vibration in the range 0-0.6Rms'g'. Primary task performance was found to vary by as much as 90% between tasks at the same Rms'g'. Differences in task difficulty accounted for this difference. Within tasks vibration-added-difficulty accounted for the effects of vibration intensity. Secondary tasks were found to be largely insensitive to vibration except secondaries which involved fine manual adjustment of minor controls. Three experiments are reported next in which an adaptive technique was used to measure the % task difficulty added by vertical random and sinusoidal vibration to a 'Critical Compensatory Tracking task. At vibration intensities between 0 - 0.09 Rms 'g' it was found that random vibration added (24.5 x Rms'g')/7.4 x 100% to the difficulty of the control task. An equivalence relationship between Random and Sinusoidal vibration effects was established based upon added task difficulty. Waveform Analyses which were applied to the experimental data served to validate Phase Plane analysis and uncovered the development of a control and possibly a vibration isolation strategy. The submission ends with an appraisal of subjects mentioned in the thesis title.

Quality control of pharmaceutical ingredients:developing a caking test for industry

Objectives - Powdered and granulated particulate materials make up most of the ingredients of pharmaceuticals and are often at risk of undergoing unwanted agglomeration, or caking, during transport or storage. This is particularly acute when bulk powders are exposed to extreme swings in temperature and relative humidity, which is now common as drugs are produced and administered in increasingly hostile climates and are stored for longer periods of time prior to use. This study explores the possibility of using a uniaxial unconfined compression test to compare the strength of caked agglomerates exposed to different temperatures and relative humidities. This is part of a longer-term study to construct a protocol to predict the caking tendency of a new bulk material from individual particle properties. The main challenge is to develop techniques that provide repeatable results yet are presented simply enough to be useful to a wide range of industries. Methods - Powdered sucrose, a major pharmaceutical ingredient, was poured into a split die and exposed to high and low relative humidity cycles at room temperature. The typical ranges were 20–30% for the lower value and 70–80% for the higher value. The outer die casing was then removed and the resultant agglomerate was subjected to an unconfined compression test using a plunger fitted to a Zwick compression tester. The force against displacement was logged so that the dynamics of failure as well as the failure load of the sample could be recorded. The experimental matrix included varying the number of cycles, the amount between the maximum and minimum relative humidity, the height and diameters of the samples, the number of cycles and the particle size. Results - Trends showed that the tensile strength of the agglomerates increased with the number of cycles and also with the more extreme swings in relative humidity. This agrees with previous work on alternative methods of measuring the tensile strength of sugar agglomerates formed from humidity cycling (Leaper et al 2003). Conclusions - The results show that at the very least the uniaxial tester is a good comparative tester to examine the caking tendency of powdered materials, with a simple arrangement and operation that are compatible with the requirements of industry. However, further work is required to continue to optimize the height/ diameter ratio during tests.

Alternative mechanisms guiding salespersons’ ambidextrous product selling

Ambidextrous product-selling strategies, in which companies’ salespeople concurrently pursue the sale of existing and new products, are hard to implement. Previous studies have addressed this issue for relatively simple consumer settings with the manager in close proximity to the salespersons and focusing on different levels of control and autonomy to resolve this issue. However, little is known about how field salespeople can be influenced to pursue such dual goals proactively for more complex business-to-business products. In this study, the authors distinguish between salespeople’s proactive selling behaviour for new and existing products, and study the impact of two alternative mechanisms: a situational mechanism (i.e. perceived manager product-selling ambidexterity) and a structural mechanism (i.e. salesperson organizational identification). Using a time-lagged, multisource data set from a large ambidextrous company, the authors demonstrate that both mechanisms contribute to salespeople’s proactive selling of new and existing products, but also act as each other’s substitutes. The results suggest two most likely strategies for salespeople to obtain overall sales targets: focusing on existing product selling; or acting ambidextrously. The latter approach offers the benefits of better achieving ambidextrous company sales goals and of greater performance stability, and is thus preferred.