Model selection approaches for non-linear system identification: a review

The identification of non-linear systems using only observed finite datasets has become a mature research area over the last two decades. A class of linear-in-the-parameter models with universal approximation capabilities have been intensively studied and widely used due to the availability of many linear-learning algorithms and their inherent convergence conditions. This article presents a systematic overview of basic research on model selection approaches for linear-in-the-parameter models. One of the fundamental problems in non-linear system identification is to find the minimal model with the best model generalisation performance from observational data only. The important concepts in achieving good model generalisation used in various non-linear system-identification algorithms are first reviewed, including Bayesian parameter regularisation and models selective criteria based on the cross validation and experimental design. A significant advance in machine learning has been the development of the support vector machine as a means for identifying kernel models based on the structural risk minimisation principle. The developments on the convex optimisation-based model construction algorithms including the support vector regression algorithms are outlined. Input selection algorithms and on-line system identification algorithms are also included in this review. Finally, some industrial applications of non-linear models are discussed.

On Variable Selection of a Nonlinear Non-parametric System with a Limited Data Set: A Stepwise Algorithm

his paper considers a problem of identification for a high dimensional nonlinear non-parametric system when only a limited data set is available. The algorithms are proposed for this purpose which exploit the relationship between the input variables and the output and further the inter-dependence of input variables so that the importance of the input variables can be established. A key to these algorithms is the non-parametric two stage input selection algorithm.

Kernel based approaches to local nonlinear non-parametric variable selection

In this paper, we consider the variable selection problem for a nonlinear non-parametric system. Two approaches are proposed, one top-down approach and one bottom-up approach. The top-down algorithm selects a variable by detecting if the corresponding partial derivative is zero or not at the point of interest. The algorithm is shown to have not only the parameter but also the set convergence. This is critical because the variable selection problem is binary, a variable is either selected or not selected. The bottom-up approach is based on the forward/backward stepwise selection which is designed to work if the data length is limited. Both approaches determine the most important variables locally and allow the unknown non-parametric nonlinear system to have different local dimensions at different points of interest. Further, two potential applications along with numerical simulations are provided to illustrate the usefulness of the proposed algorithms.

Comparative genomic analysis reveals species-dependent complexities that explain difficulties with microsatellite marker development in molluscs

Reliable population DNA molecular markers are difficult to develop for molluscs, the reasons for which are largely unknown. Identical protocols for microsatellite marker development were implemented in three gastropods. Success rates were lower for Gibbula cineraria compared to Littorina littorea and L. saxatilis. Comparative genomic analysis of 47.2?kb of microsatellite containing sequences (MCS) revealed a high incidence of cryptic repetitive DNA in their flanking regions. The majority of these were novel, and could be grouped into DNA families based upon sequence similarities. Significant inter-specific variation in abundance of cryptic repetitive DNA and DNA families was observed. Repbase scans show that a large proportion of cryptic repetitive DNA was identified as transposable elements (TEs). We argue that a large number of TEs and their transpositional activity may be linked to differential rates of DNA multiplication and recombination. This is likely to be an important factor explaining inter-specific variation in genome stability and hence microsatellite marker development success rates. Gastropods also differed significantly in the type of TEs classes (autonomous vs non-autonomous) observed. We propose that dissimilar transpositional mechanisms differentiate the TE classes in terms of their propensity for transposition, fixation and/or silencing. Consequently, the phylogenetic conservation of non-autonomous TEs, such as CvA, suggests that dispersal of these elements may have behaved as microsatellite-inducing elements. Results seem to indicate that, compared to autonomous, non-autonomous TEs maybe have a more active role in genome rearrangement processes. The implications of the findings for genomic rearrangement, stability and marker development are discussed.

Inactivation of Rb in stromal fibroblasts promotes epithelial cell invasion

Stromal-derived growth factors are required for normal epithelial growth but are also implicated in tumour progression. We have observed inactivation of the retinoblastoma protein (Rb), through phosphorylation, in cancer-associated fibroblasts in oro-pharyngeal cancer specimens. Rb is well known for its cell-autonomous effects on cancer initiation and progression; however, cell non-autonomous functions of Rb are not well described. We have identified a cell non-autonomous role of Rb, using three-dimensional cultures, where depletion of Rb in stromal fibroblasts enhances invasive potential of transformed epithelia. In part, this is mediated by upregulation of keratinocyte growth factor (KGF), which is produced by the depleted fibroblasts. KGF drives invasion of epithelial cells through induction of MMP1 expression in an AKT- and Ets2-dependent manner. Our data identify that stromal fibroblasts can alter the invasive behaviour of the epithelium, and we show that altered expression of KGF can mediate these functions. © European Molecular Biology Organization.

Day-Ahead Optimal Charging/Discharging Scheduling for Electric Vehicles in Micro-Grids

A micro-grid is an autonomous system which can be operated and connected to an external system or isolated with the help of energy storage systems (ESSs). While the daily output of distributed generators (DGs) strongly depends on the temporal distribution of natural resources such as wind and solar, unregulated electric vehicle (EV) charging demand will deteriorate the imbalance between the daily load and generation curves. In this paper, a statistical model is presented to describe daily EV charging/discharging behaviour. An optimisation problem is proposed to obtain economic operation for the micro-grid based on this model. In day-ahead scheduling, with estimated information of power generation and load demand, optimal charging/discharging of EVs during 24 hours is obtained. A series of numerical optimization solutions in different scenarios is achieved by serial quadratic programming. The results show that optimal charging/discharging of EVs, a daily load curve can better track the generation curve and the network loss and required ESS capacity are both decreased. The paper also demonstrates cost benefits for EVs and operators.

Characterization of a thymidylate synthase (TS)-inducible cell line: a model system for studying sensitivity to TS- and non-TS-targeted chemotherapies

Thymidylate synthase (TS) is responsible for the de novo synthesis of thymidylate, which is required for DNA synthesis and repair and which is an important target for fluoropyrimidines such as 5-fluorouracil (5-FU), and antifolates such as Tomudex (TDX), ZD9331, and multitargeted antifolate (MTA). To study the importance of TS expression in determining resistance to these agents, we have developed an MDA435 breast cancer-derived cell line with tetracycline-regulated expression of TS termed MTS-5. We have demonstrated that inducible expression of TS increased the IC(50) dose of the TS-targeted therapeutic agents 5-FU, TDX, and ZD9331 by 2-, 9- and 24-fold respectively. An IC(50) dose for MTA was unobtainable when TS was overexpressed in these cells, which indicated that MTA toxicity is highly sensitive to increased TS expression levels. The growth inhibitory effects of the chemotherapeutic agents CPT-11, cisplatin, oxaliplatin, and Taxol were unaffected by TS up-regulation. Cell cycle analyses revealed that IC(50) doses of 5-FU, TDX and MTA caused an S-phase arrest in cells that did not overexpress TS, and this arrest was overcome when TS was up-regulated. Furthermore, the S-phase arrest was accompanied by 2- to 4-fold increased expression of the cell cycle regulatory genes cyclin E, cyclin A, and cyclin dependent kinase 2 (cdk2). These results indicate that acute increases in TS expression levels play a key role in determining cellular sensitivity to TS-directed chemotherapeutic drugs by modulating the degree of S-phase arrest caused by these agents. Moreover, CPT-11, cisplatin, oxaliplatin, and Taxol remain highly cytotoxic in cells that overexpress TS.

An autonomous wave-powered desalination system

The potential for an autonomous wave-powered desalination system is considered and it is identified that the most promising configuration is a reverse osmosis (RO) plant utilising a pressure exchanger-intensifier for energy recovery. A numerical model of the RO plant with a pressure exchanger-intensifier is developed that shows that a specific energy consumption of less than 2.0 kW h/m3 over a wide range of sea-water feed conditions, making it particularly suitable for use with a variable power source such as wave energy. A numerical model of the combined wave-power and desalination plant is also developed that shows that it is possible to supply the desalination plant with sea-water directly pressurised by the wave energy converter, eliminating the cost and energy losses associated with converting the energy into electricity and back to pressurised water. For a typical sea-state the specific hydraulic energy consumption of the desalination plant is estimated to be 1.85 kW h/m3 whilst maintaining a recovery-ratio of less than 25 to 35% to avoid the need for chemical pre-treatment to eliminate scaling problems. It is suggested that the economic potential for wave-powered desalination depends on these energy and cost savings more than compensating for the reduction in membrane life that occurs with variable feed conditions.

Building Autonomous Sensitive Artificial Listeners

This paper describes a substantial effort to build a real-time interactive multimodal dialogue system with a focus on emotional and non-verbal interaction capabilities. The work is motivated by the aim to provide technology with competences in perceiving and producing the emotional and non-verbal behaviours required to sustain a conversational dialogue. We present the Sensitive Artificial Listener (SAL) scenario as a setting which seems particularly suited for the study of emotional and non- verbal behaviour, since it requires only very limited verbal understanding on the part of the machine. This scenario allows us to concentrate on non-verbal capabilities without having to address at the same time the challenges of spoken language understanding, task modeling etc. We first report on three prototype versions of the SAL scenario, in which the behaviour of the Sensitive Artificial Listener characters was determined by a human operator. These prototypes served the purpose of verifying the effectiveness of the SAL scenario and allowed us to collect data required for building system components for analysing and synthesising the respective behaviours. We then describe the fully autonomous integrated real-time system we created, which combines incremental analysis of user behaviour, dialogue management, and synthesis of speaker and listener behaviour of a SAL character displayed as a virtual agent. We discuss principles that should underlie the evaluation of SAL-type systems. Since the system is designed for modularity and reuse, and since it is publicly available, the SAL system has potential as a joint research tool in the affective computing research community.

Non-aqueous silicone elastomer gels as a vaginal microbicide delivery system for the HIV-1 entry inhibitor maraviroc

Aqueous semi-solid polymeric gels, such as those based on hydroxyethylcellulose (HEC) and polyacrylic acid (e.g. Carbopol®), have a long history of use in vaginal drug delivery. However, despite their ubiquity, they often provide sub-optimal clinical performance, due to poor mucosal retention and limited solubility for poorly water-soluble actives. These issues are particularly pertinent for vaginal HIV microbicides, since many lead candidates are poorly water-soluble and where a major goal is the development of a coitally independent, once daily gel product. In this study, we report the use of a non-aqueous silicone elastomer gel for vaginal delivery of the HIV-1 entry inhibitor maraviroc. In vitro rheological, syringeability and retention studies demonstrated enhanced performance for silicone gels compared with a conventional aqueous HEC gel, while testing of the gels in the slug model confirmed a lack of mucosal irritancy. Pharmacokinetic studies following single dose vaginal administration of a maraviroc silicone gel in rhesus macaques showed higher and sustained MVC levels in vaginal fluid, vaginal tissue and plasma compared with a HEC gel containing the same maraviroc loading. The results demonstrate that non-aqueous silicone gels have potential as a formulation platform for coitally independent vaginal HIV microbicides.

Role of environmental correlations in the non-Markovian dynamics of a spin system

We study the dynamics of a chain of interacting quantum particles affected by an individual or collective environment(s), focusing on the role played by the environmental quantum correlations over the evolution of the chain. The presence of entanglement in the state of the environment magnifies the non-Markovian nature of the chain's dynamics, giving rise to structures in figures of merit such as spin entanglement and purity that are not observed under a separable environmental state. Our analysis can be relevant to problems tackling the open-system dynamics of biological complexes of strong current interest.