420 research outputs found
Bone Marrow Stem Cell Treatment for Ischemic Heart Disease in Patients with No Option of Revascularization: A Systematic Review and Meta-Analysis
No full textPMCID: PMC3686792This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Classical and novel hormonal influences on renal tubular transport and the emrging concept of intracrine regulation
No full textA systematic review and economic evaluation of epoetin alpha,epoetin beta and darbepoetin alpha in anaemia associated with cancer, especially that attributable to cancer treatment
No full textObjectives: to assess the effectiveness and cost-effectiveness of epoetin alpha, epoetin beta and darbepoetin alpha (referred to collectively in this report as epo) in anaemia associated with cancer, especially that attributable to cancer treatment.Data sources: electronic databases were searched from 2000 (1996 in the case of darbepoetin alpha) to September 2004.Review methods: using a recently published Cochrane review as the starting point, a systematic review of recent randomised controlled trials (RCTs) comparing epo with best standard was conducted. Inclusion, quality assessment and data abstraction were undertaken in duplicate. Where possible, meta-analysis was employed. The economic assessment consisted of a systematic review of past economic evaluations, an assessment of economic models submitted by the manufacturers of the three epo agents and development of a new individual sampling model (the Birmingham epo model).Results: in total 46 RCTs were included within this systematic review, 27 of which had been included in the Cochrane systematic review. All 46 trials compared epo plus supportive care for anaemia (including transfusions), with supportive care for anaemia (including transfusions), alone. Haematological response (defined as an improvement by 2 g/dl(-1)) had a relative risk of 3.4 [95% confidence interval (CI) 3.0 to 3.8, 22 RCTs] with a response rate for epo of 53%. The trial duration was most commonly 16-20 weeks. There was little statistical heterogeneity in the estimate of haematological response, and there were no important differences between the subgroups examined. Haemoglobin (Hb) change showed a weighted mean difference of 1.63 g/dl(-1) (95% CI 1.46 to 1.80) in favour of epo. Treatment with erythropoietin in patients with cancer-induced anaemia reduces the number of patients who receive a red blood cell transfusion (RBCT) by an estimated 18%. Health-related quality of life (HRQoL) data were analysed using vote counting and qualitative assessment and a positive effect was observed in favour of an improved HRQoL for patients on epo. Published information on side-effects was of poor quality. New trials provided further evidence of side-effects with epo, particularly thrombic events, but it is still unclear whether these could be accounted for by chance alone. The results of the previous Cochrane review had suggested a survival advantage for epo (HR 0.84, 95% CI 0.69 to 1.02), based on 19 RCTs. The update, based on 28 RCTs, suggests no difference (HR 1.03, 95% CI 0.88 to 1.21). Subgroup analysis suggested some explanations for this heterogeneity, but it is difficult to draw firm conclusions without access to the substantial amounts of missing or unpublished data, or more detailed results from some of the trials with heterogeneous patient populations. The conclusions are, however, broadly in line with those of a Food and Drug Administration (FDA) safety briefing, which recommended that patients with a haemoglobin above 12 g/dl(-1) should not be treated; the target rate of rise in Hb should not be too great, and further carefully conducted trials are required to determine which subgroups of patients may be harmed by the use of these products, in particular through the stimulation of tumour activity. Five published economic evaluations identified from the literature had inconsistent results, with estimates ranging from a cost per quality-adjusted life-year (QALY) under pound 10,000 through to epo being less effective and more costly than standard care. The more favourable evaluations assumed a survival advantage for epo. The three company models submitted each relied on assumed survival gains to achieve relatively low cost per QALY, from pound 13,000 to pound 28,000, but generated estimates from pound 84,000 to pound 159,000 per QALY when no survival gain was assumed. Each of these models relied on Hb levels alone driving utility, and each assumed gradual normalisation of Hb in the standard treatment arm after the end of treatment. The Birmingham epo model followed the company models in regard to the relationship between Hb levels and utility, and also assumed normalisation in the base case. With no survival gain, the incremental cost per QALY was pound 150,000, falling to pound 40,000 when the lower, more favourable, confidence interval for survival was used.Conclusions: epo is effective in improving haematological response and reducing RBCT requirements, and appears to have a positive effect on HRQoL. The incidence of side-effects and effects on survival remains highly uncertain. However, if there is no impact on survival, it seems highly unlikely that epo would be considered a cost-effective use of healthcare resources. The main target for further research should be improving estimates of impact on survival, initially through more detailed secondary research, such as the individual patient data meta-analysis started by the Cochrane group. Further trials may be required, and have been recommended by the FDA, although many trials are in progress, completed but unreported or awaiting mature follow-up. The Birmingham epo model developed as part of this project contains new features that improve its flexibility in exploring different scenarios; further refinement and validation would therefore be of assistance. Finally, further research to resolve uncertainty about other parameters, particularly quality of life, adverse events, and the rate of normalisation, would also be beneficia
BNN-DP: Robustness Certification of Bayesian Neural Networks via Dynamic Programming
No full textIn this paper, we introduce BNN-DP, an efficient algorithmic framework for analysis of adversarial robustness of Bayesian Neural Networks (BNNs). Given a compact set of input points T ⊂ Rn, BNN-DP computes lower and upper bounds on the BNN's predictions for all the points in T. The framework is based on an interpretation of BNNs as stochastic dynamical systems, which enables the use of Dynamic Programming (DP) algorithms to bound the prediction range along the layers of the network. Specifically, the method uses bound propagation techniques and convex relaxations to derive a backward recursion procedure to over-approximate the prediction range of the BNN with piecewise affine functions. The algorithm is general and can handle both regression and classification tasks. On a set of experiments on various regression and classification tasks and BNN architectures, we show that BNN-DP outperforms state-of-the-art methods by up to four orders of magnitude in both tightness of the bounds and computational efficiency.Team Luca Laurent
The exact s-matrices of affine toda field theories
No full textThis thesis is concerned with exact solutions to various massive field theories in 1+1 dimensions.Two approaches are described. The first, abstract and non-lagrangian, relies on the considerable understanding that there now is of massless two-dimensional field theories. A perturbative scheme can be developed within which various exact statements may be made. Chapter 1 contains a review of this technique, together with some work applying it in various simple situations. The particular structures studied turn out to have a deep connection with certain Lie algebras, a fact which is discussed in the concluding three sections of the chapter. A complementary approach is to study specific, classically integrable, lagrangians in the hope that their quantum versions will also permit an exact treatment. Motivated to some extent by the findings of chapter 1, the remainder of the thesis is devoted to a particular class of models known as affine Toda field theories. Mixtures of perturbative and non-perturbative ideas are employed. The non-perturbative elements are to be found in analytic S-matrix theory, reviewed in chapter 2, while various features of the classical theory necessary for a perturbative quantum treatment are derived in chapter 3. Making use of this information, chapter 4 proposes exact expressions for the S-matrices for a large subset of the Toda theories, which are then checked in perturbation theory. Finally, the relevance or otherwise of the Toda S-matrices to the perturbations of massless theories studied in chapter 1 is discussed, and some possible directions for future work are mentioned
The electronics industry: inward investment versus indigenous development -- the policy debate
No full textIn this paper the public policy implications of an active government strategy aimed at enhancing the competitiveness of the electronics industry in Britain are examined. The author argues that as a general principle industrial policy should be both designed and applied at as low a level as possible. To achieve this a comprehensive but decentralised institutional economic development network will need to be created.
Experimental and numerical modelling of wheel rail contact and wear
No full textIn the field of simulation of railroad vehicles, there are many numerical models to estimate the interaction forces between the wheel and rail. The main advantage of these models is that they can be used together with multi-body dynamics software to calculate the motion of a vehicle in real time. However, the result of these contact models is usually post-processed to estimate wear on the profiles and some hypotheses assumed by the contact models may be inadequate for wear analysis. This is the case when considering surface roughness, which is not introduced in the numerical models and makes wear prediction imprecise. In this work an experimental method based on the measurement of ultrasonic reflection is used to solve the contact problem, together with a FASTSIM (simplified theory of rolling contact) algorithm. This technique is suitable to deal with rough surfaces and gives a better approximation of the material behaviour. Wear is estimated by means of the energy dissipation approach (T·gamma). Two different contacts are investigated, using wheel and rail profiles coming from unused and worn specimens. In order to obtain realistic results, special care is taken when locating the specimens to reproduce the same contact that appears between the wheel and the rail in the track.The corresponding author gratefully acknowledges the cooperation of C. Hardwick and Portec Rail Inc. for supplying the Miniprof device. This research was supported by Universitat Politecnica de Valencia (Spain).Rovira Cardete, A.; Roda Buch, A.; Marshall, M.; Brunskill, H.; Lewis, R. (2011). Experimental and numerical modelling of wheel rail contact and wear. Wear. 271(5-6):911-924. doi:10.1016/j.wear.2011.03.024S9119242715-
Trading-Off Payments and Accuracy in Online Classification with Paid Stochastic Experts
No full textWe investigate online classification with paid stochastic experts. Here, before making their prediction, each expert must be paid. The amount that we pay each expert directly influences the accuracy of their prediction through some unknown Lipschitz “productivity” function. In each round, the learner must decide how much to pay each expert and then make a prediction. They incur a cost equal to a weighted sum of the prediction error and upfront payments for all experts. We introduce an online learning algorithm whose total cost after rounds exceeds that of a predictor which knows the productivity of all experts in advance by at most where is the number of experts. In order to achieve this result, we combine Lipschitz bandits and online classification with surrogate losses. These tools allow us to improve upon the bound of order one would obtain in the standard Lipschitz bandit setting. Our algorithm is empirically evaluated on synthetic data
Towards Theoretical Understanding of Inverse Reinforcement Learning
No full textInverse reinforcement learning (IRL) denotes a powerful family of algorithms for recovering a reward function justifying the behavior demonstrated by an expert agent. A well-known limitation of IRL is the ambiguity in the choice of the reward function, due to the existence of multiple rewards that explain the observed behavior. This limitation has been recently circumvented by formulating IRL as the problem of estimating the feasible reward set, i.e., the region of the rewards compatible with the expert’s behavior. In this paper, we make a step towards closing the theory gap of IRL in the case of finite-horizon problems with a generative model. We start by formally introducing the problem of estimating the feasible reward set, the corresponding PAC requirement, and discussing the properties of particular classes of rewards. Then, we provide the first minimax lower bound on the sample complexity for the problem of estimating the feasible reward set of order , being and the number of states and actions respectively, the horizon, the desired accuracy, and the confidence. We analyze the sample complexity of a uniform sampling strategy (US-IRL), proving a matching upper bound up to logarithmic factors. Finally, we outline several open questions in IRL and propose future research directions
Optimally-weighted Estimators of the Maximum Mean Discrepancy for Likelihood-Free Inference
No full textLikelihood-free inference methods typically make use of a distance between simulated and real data. A common example is the maximum mean discrepancy (MMD), which has previously been used for approximate Bayesian computation, minimum distance estimation, generalised Bayesian inference, and within the nonparametric learning framework. The MMD is commonly estimated at a root-m rate, where m is the number of simulated samples. This can lead to significant computational challenges since a large m is required to obtain an accurate estimate, which is crucial for parameter estimation. In this paper, we propose a novel estimator for the MMD with significantly improved sample complexity. The estimator is particularly well suited for computationally expensive smooth simulators with low- to mid-dimensional inputs. This claim is supported through both theoretical results and an extensive simulation study on benchmark simulators
- …
