Mixture density network training by computation in parameter space

Training Mixture Density Network (MDN) configurations within the NETLAB framework takes time due to the nature of the computation of the error function and the gradient of the error function. By optimising the computation of these functions, so that gradient information is computed in parameter space, training time is decreased by at least a factor of sixty for the example given. Decreased training time increases the spectrum of problems to which MDNs can be practically applied making the MDN framework an attractive method to the applied problem solver.

CFD modelling of the fast pyrolysis of biomass in fluidised bed reactors, Part A: Eulerian computation of momentum transport in bubbling fluidised beds

The fluid–particle interaction inside a 150 g/h fluidised bed reactor is modelled. The biomass particle is injected into the fluidised bed and the momentum transport from the fluidising gas and fluidised sand is modelled. The Eulerian approach is used to model the bubbling behaviour of the sand, which is treated as a continuum. The particle motion inside the reactor is computed using drag laws, dependent on the local volume fraction of each phase, according to the literature. FLUENT 6.2 has been used as the modelling framework of the simulations with a completely revised drag model, in the form of user defined function (UDF), to calculate the forces exerted on the particle as well as its velocity components. 2-D and 3-D simulations are tested and compared. The study is the first part of a complete pyrolysis model in fluidised bed reactors.

Protecting agents against malicious host attacks

The introduction of agent technology raises several security issues that are beyond conventional security mechanisms capability and considerations, but research in protecting the agent from malicious host attack is evolving. This research proposes two approaches to protecting an agent from being attacked by a malicious host. The first approach consists of an obfuscation algorithm that is able to protect the confidentiality of an agent and make it more difficult for a malicious host to spy on the agent. The algorithm uses multiple polynomial functions with multiple random inputs to convert an agent's critical data to a value that is meaningless to the malicious host. The effectiveness of the obfuscation algorithm is enhanced by addition of noise code. The second approach consists of a mechanism that is able to protect the integrity of the agent using state information, recorded during the agent execution process in a remote host environment, to detect a manipulation attack by a malicious host. Both approaches are implemented using a master-slave agent architecture that operates on a distributed migration pattern. Two sets of experimental test were conducted. The first set of experiments measures the migration and migration+computation overheads of the itinerary and distributed migration patterns. The second set of experiments is used to measure the security overhead of the proposed approaches. The protection of the agent is assessed by analysis of its effectiveness under known attacks. Finally, an agent-based application, known as Secure Flight Finder Agent-based System (SecureFAS) is developed, in order to prove the function of the proposed approaches.

The SECURE study:long-term safety of Ranibizumab 0.5 mg in neovascular age-related macular degeneration

Objective - To evaluate long-term safety of intravitreal ranibizumab 0.5-mg injections in neovascular age-related macular degeneration (nAMD). Design - Twenty-four–month, open-label, multicenter, phase IV extension study. Participants - Two hundred thirty-four patients previously treated with ranibizumab for 12 months in the EXCITE/SUSTAIN study. Methods - Ranibizumab 0.5 mg administered at the investigator's discretion as per the European summary of product characteristics 2007 (SmPC, i.e., ranibizumab was administered if a patient experienced a best-corrected visual acuity [BCVA] loss of >5 Early Treatment Diabetic Retinopathy Study letters measured against the highest visual acuity [VA] value obtained in SECURE or previous studies [EXCITE and SUSTAIN], attributable to the presence or progression of active nAMD in the investigator's opinion). Main Outcome Measures - Incidence of ocular or nonocular adverse events (AEs) and serious AEs, mean change in BCVA from baseline over time, and the number of injections. Results - Of 234 enrolled patients, 210 (89.7%) completed the study. Patients received 6.1 (mean) ranibizumab injections over 24 months. Approximately 42% of patients had 7 or more visits at which ranibizumab was not administered, although they had experienced a VA loss of more than 5 letters, indicating either an undertreatment or that factors other than VA loss were considered for retreatment decision by the investigator. The most frequent ocular AEs (study eye) were retinal hemorrhage (12.8%; 1 event related to study drug), cataract (11.5%; 1 event related to treatment procedure), and increased intraocular pressure (6.4%; 1 event related to study drug). Cataract reported as serious due to hospitalization for cataract surgery occurred in 2.6% of patients; none was suspected to be related to study drug or procedure. Main nonocular AEs were hypertension and nasopharyngitis (9.0% each). Arterial thromboembolic events were reported in 5.6% of the patients. Five (2.1%) deaths occurred during the study, none related to the study drug or procedure. At month 24, mean BCVA declined by 4.3 letters from the SECURE baseline. Conclusions - The SECURE study showed that ranibizumab administered as per a VA-guided flexible dosing regimen recommended in the European ranibizumab SmPC at the investigator's discretion was well tolerated over 2 years. No new safety signals were identified in patients who received ranibizumab for a total of 3 years. On average, patients lost BCVA from the SECURE study baseline, which may be the result of disease progression or possible undertreatment.

Analogue computation of the impedence of a powered flying control system

DUE TO COPYRIGHT RESTRICTIONS ONLY AVAILABLE FOR CONSULTATION AT ASTON UNIVERSITY LIBRARY AND INFORMATION SERVICES WITH PRIOR ARRANGEMENT

A data structure for improved gp analysis via efficient computation and visualisation of population measures

Population measures for genetic programs are defined and analysed in an attempt to better understand the behaviour of genetic programming. Some measures are simple, but do not provide sufficient insight. The more meaningful ones are complex and take extra computation time. Here we present a unified view on the computation of population measures through an information hypertree (iTree). The iTree allows for a unified and efficient calculation of population measures via a basic tree traversal. © Springer-Verlag 2004.

Secure key distribution over a 500 km long link using a Raman ultra-long fiber laser

In traditional communication systems the transmission medium is considered as a given characteristic of the channel, which does not depend on the properties of the transmitter and the receiver. Recent experimental demonstrations of the feasibility of extending the laser cavity over the whole communication link connecting the two parties, forming an ultra-long fiber laser (UFL), have raised groundbreaking possibilities in communication and particularly in secure communications. Here, a 500 km long secure key distribution link based on Raman gain UFL is demonstrated. An error-free distribution of a random key with an average rate of 100 bps between the users is demonstrated and the key is shown to be unrecoverable to an eavesdropper employing either time or frequency domain passive attacks. In traditional communication systems the transmission medium is considered as a given characteristic of the channel, which does not depend on the properties of the transmitter and the receiver. Recent demonstrations of the feasibility of extending the laser cavity over the whole communication link connecting the two parties, forming an ultra-long fiber laser (UFL), have raised groundbreaking possibilities in communication. Here, a 500 km long secure key distribution link based on Raman gain UFL is demonstrated. © 2014 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Developments towards practical free-space quantum cryptography

We describe a free space quantum cryptography system which is designed to allow continuous unattended key exchanges for periods of several days, and over ranges of a few kilometres. The system uses a four-laser faint-pulse transmission system running at a pulse rate of 10MHz to generate the required four alternative polarization states. The receiver module similarly automatically selects a measurement basis and performs polarization measurements with four avalanche photodiodes. The controlling software can implement the full key exchange including sifting, error correction, and privacy amplification required to generate a secure key.

On reliable computation by noisy random Boolean formulas

We study noisy computation in randomly generated k-ary Boolean formulas. We establish bounds on the noise level above which the results of computation by random formulas are not reliable. This bound is saturated by formulas constructed from a single majority-like gate. We show that these gates can be used to compute any Boolean function reliably below the noise bound.

DNA Simulation of Genetic Algorithms: Fitness Computation

* This work has been partially supported by Spanish Project TIC2003-9319-c03-03 “Neural Networks and Networks of Evolutionary Processors”.

Generalization by Computation Through Memory

Usually, generalization is considered as a function of learning from a set of examples. In present work on the basis of recent neural network assembly memory model (NNAMM), a biologically plausible 'grandmother' model for vision, where each separate memory unit itself can generalize, has been proposed. For such a generalization by computation through memory, analytical formulae and numerical procedure are found to calculate exactly the perfectly learned memory unit's generalization ability. The model's memory has complex hierarchical structure, can be learned from one example by a one-step process, and may be considered as a semi-representational one. A simple binary neural network for bell-shaped tuning is described.

On the Error-Free Computation of Fast Cosine Transform

We extend our previous work into error-free representations of transform basis functions by presenting a novel error-free encoding scheme for the fast implementation of a Linzer-Feig Fast Cosine Transform (FCT) and its inverse. We discuss an 8x8 L-F scaled Discrete Cosine Transform where the architecture uses a new algebraic integer quantization of the 1-D radix-8 DCT that allows the separable computation of a 2-D DCT without any intermediate number representation conversions. The resulting architecture is very regular and reduces latency by 50% compared to a previous error-free design, with virtually the same hardware cost.

Practical, Computation Efficient High-Order Neural Network for Rotation and Shift Invariant Pattern Recognition

In this paper, a modification for the high-order neural network (HONN) is presented. Third order networks are considered for achieving translation, rotation and scale invariant pattern recognition. They require however much storage and computation power for the task. The proposed modified HONN takes into account a priori knowledge of the binary patterns that have to be learned, achieving significant gain in computation time and memory requirements. This modification enables the efficient computation of HONNs for image fields of greater that 100 × 100 pixels without any loss of pattern information.

Experimental Support of Argument-based Syntactic Computation

Linguistic theory, cognitive, information, and mathematical modeling are all useful while we attempt to achieve a better understanding of the Language Faculty (LF). This cross-disciplinary approach will eventually lead to the identification of the key principles applicable in the systems of Natural Language Processing. The present work concentrates on the syntax-semantics interface. We start from recursive definitions and application of optimization principles, and gradually develop a formal model of syntactic operations. The result – a Fibonacci- like syntactic tree – is in fact an argument-based variant of the natural language syntax. This representation (argument-centered model, ACM) is derived by a recursive calculus that generates a mode which connects arguments and expresses relations between them. The reiterative operation assigns primary role to entities as the key components of syntactic structure. We provide experimental evidence in support of the argument-based model. We also show that mental computation of syntax is influenced by the inter-conceptual relations between the images of entities in a semantic space.