Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Networks”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 415 records · Page 23Linked to original sources
PubMed2005

[Studies of gene networks and singal transduction of kidney deficiency syndrome by syndrome differentiation through drug effects].

Three kinds of networks were summarily described in this review including the small intracellular molecular networks, the middle-scale networks of hypothalamus-pituitary-adrenal-thymus (HPAT) axis and the large network, neuroendocrine-immune (NEI) network, covering the whole organism and linking multiple systems together. The hypothesis was expressed that the "disease" or "syndrome" formed in the human body by the intervention from outside world is based on the changes of multi-molecular network. In this paper, the pattern and ability of signal transduction channel and the methods of studying changes in it were also described, and raised, herefrom, "to determine syndrome by drug effects (DSDE)" is the intervention means for studying syndrome in the light of systemic biological methods. We found Kidney-yang deficiency syndrome covered the NEI network and the regulating center located in hypothamus with Compound Bushen Recipe (CBR, Kidney-tonifying recipe). By intervention with EF, an effective component of CBR, it was found that EF can activate the immune system and the three networks, including growth axis, sex hormone axis and lymphocyte apoptosis network in HPAT axis through the downward pathway of NEI network to play its efficiency of molecular network. There are many regulation patterns of EF on networks. For example, in the network mechanism of lymphocyte apoptosis and proliferation, EF can reconstruct the balance of the opposite apoptosis related genes and proliferation related genes; EF can assemble and integrate co-stimulating molecules, transform growth factors (TGF), and several oncogenes to form an upstream factor network for initiating the proliferation and anti-apoptosis promotion; EF can simultaneously up-regulate the two opposite genes expression of IkappaB and NFkappaB in NIK/IKK/IkappaB/Rel/NFkappaB signal transduction channel, which could not only control the rising of NFkappaB in a moderate range, but also guarantee its predominant status to exert its hinge role in molecule regulating network, by which gene network regulation atlas in HPAT axis of Kidney-deficiency syndrome was observed.

Animals↗
PubMed2002

On the complexity of computing and learning with multiplicative neural networks.

In a great variety of neuron models, neural inputs are combined using the summing operation. We introduce the concept of multiplicative neural networks that contain units that multiply their inputs instead of summing them and thus allow inputs to interact nonlinearly. The class of multiplicative neural networks comprises such widely known and well-studied network types as higher-order networks and product unit networks. We investigate the complexity of computing and learning for multiplicative neural networks. In particular, we derive upper and lower bounds on the Vapnik-Chervonenkis (VC) dimension and the pseudo-dimension for various types of networks with multiplicative units. As the most general case, we consider feedforward networks consisting of product and sigmoidal units, showing that their pseudo-dimension is bounded from above by a polynomial with the same order of magnitude as the currently best-known bound for purely sigmoidal networks. Moreover, we show that this bound holds even when the unit type, product or sigmoidal, may be learned. Crucial for these results are calculations of solution set components bounds for new network classes. As to lower bounds, we construct product unit networks of fixed depth with super-linear VC dimension. For sigmoidal networks of higher order, we establish polynomial bounds that, in contrast to previous results, do not involve any restriction of the network order. We further consider various classes of higher-order units, also known as sigma-pi units, that are characterized by connectivity constraints. In terms of these, we derive some asymptotically tight bounds. Multiplication plays an important role in both neural modeling of biological behavior and computing and learning with artificial neural networks. We briefly survey research in biology and in applications where multiplication is considered an essential computational element. The results we present here provide new tools for assessing the impact of multiplication on the computational power and the learning capabilities of neural networks.

Artificial Intelligence↗
PubMed2003

Network capacity analysis for latent attractor computation.

Attractor networks have been one of the most successful paradigms in neural computation, and have been used as models of computation in the nervous system. Recently, we proposed a paradigm called 'latent attractors' where attractors embedded in a recurrent network via Hebbian learning are used to channel network response to external input rather than becoming manifest themselves. This allows the network to generate context-sensitive internal codes in complex situations. Latent attractors are particularly helpful in explaining computations within the hippocampus--a brain region of fundamental significance for memory and spatial learning. Latent attractor networks are a special case of associative memory networks. The model studied here consists of a two-layer recurrent network with attractors stored in the recurrent connections using a clipped Hebbian learning rule. The firing in both layers is competitive--K winners take all firing. The number of neurons allowed to fire, K, is smaller than the size of the active set of the stored attractors. The performance of latent attractor networks depends on the number of such attractors that a network can sustain. In this paper, we use signal-to-noise methods developed for standard associative memory networks to do a theoretical and computational analysis of the capacity and dynamics of latent attractor networks. This is an important first step in making latent attractors a viable tool in the repertoire of neural computation. The method developed here leads to numerical estimates of capacity limits and dynamics of latent attractor networks. The technique represents a general approach to analyse standard associative memory networks with competitive firing. The theoretical analysis is based on estimates of the dendritic sum distributions using Gaussian approximation. Because of the competitive firing property, the capacity results are estimated only numerically by iteratively computing the probability of erroneous firings. The analysis contains two cases: the simple case analysis which accounts for the correlations between weights due to shared patterns and the detailed case analysis which includes also the temporal correlations between the network's present and previous state. The latter case predicts better the dynamics of the network state for non-zero initial spurious firing. The theoretical analysis also shows the influence of the main parameters of the model on the storage capacity.

Animals↗
PubMed2002

Mathematical modelling of flow through vascular networks: implications for tumour-induced angiogenesis and chemotherapy strategies.

Angiogenesis, the formation of blood vessels from a pre-existing vasculature, is a process whereby capillary sprouts are formed in response to externally supplied chemical stimuli. The sprouts then grow and develop, driven initially by endothelial cell migration, and organize themselves into a branched, connected network structure. Subsequent cell proliferation near the sprout-tip permits further extension of the capillary and ultimately completes the process. Angiogenesis occurs during embryogenesis, wound healing, arthritis and during the growth of solid tumours. In this paper we initially generate theoretical capillary networks (which are morphologically similar to those networks observed in vivo) using the discrete mathematical model of Anderson and Chaplain. This discrete model describes the formation of a capillary sprout network via endothelial cell migratory and proliferative responses to external chemical stimuli (tumour angiogenic factors, TAF) supplied by a nearby solid tumour, and also the endothelial cell interactions with the extracellular matrix. The main aim of this paper is to extend this work to examine fluid flow through these theoretical network structures. In order to achieve this we make use of flow modelling tools and techniques (specifically, flow through interconnected networks) from the field of petroleum engineering. Having modelled the flow of a basic fluid through our network, we then examine the effects of fluid viscosity, blood vessel size (i.e., diameter of the capillaries), and network structure/geometry, upon: (i) the rate of flow through the network; (ii) the amount of fluid present in the complete network at any one time; and (iii) the amount of fluid reaching the tumour. The incorporation of fluid flow through the generated vascular networks has highlighted issues that may have major implications for the study of nutrient supply to the tumour (blood/oxygen supply) and, more importantly, for the delivery of chemotherapeutic drugs to the tumour. Indeed, there are also implications for the delivery of anti-angiogenesis drugs to the network itself. Results clearly highlight the important roles played by the structure and morphology of the network, which is, in turn, linked to the size and geometry of the nearby tumour. The connectedness of the network, as measured by the number of loops formed in the network (the anastomosis density), is also found to be of primary significance. Moreover, under certain conditions, the results of our flow simulations show that an injected chemotherapy drug may bypass the tumour altogether.

Angiogenesis Inducing Agents↗
PubMed2006

Mechanisms through which drug, sex partner, and friendship network characteristics relate to risky needle use among high risk youth and young adults.

Drug injector social networks are a primary social space in which risky needle use behaviors associated with HIV transmission occur, but the mechanisms through which these social networks influence risky needle use are unclear. This study investigated two mechanisms, social support and social regulation, through which injection drug users' social networks might relate to risky needle use behaviors. We investigated how these mechanisms work in three types of social networks, namely, drug user, sex partner, and friendship networks. Data are from a study of HIV risk and protective behaviors of youth and young adults, ages 14-43, in the United States who were injection drug users and/or the sexual partners of users (N = 277). The three types of networks were constructed based on information respondents provided about their drug use partners, sexual partners, and friends. The networks were characterized by structural (i.e., size and density) and interactional (i.e., multiplexity and closeness) characteristics. We conducted tests for mediation using ordered probit models and multiple linear regression. In the drug networks, social regulation partially meditated the relationship between multiplexity and risky needle use (p<.10). In the sex partner networks, emotional support partially mediated the relationship between network size and risky needle use (p<.001), fully mediated the relationship between network closeness and risky needle use (p<.001), and partially mediated the relationship between multiplexity and risky needle use (p<.01). Mediators were not identified in the friendship networks. The findings demonstrate that network mechanisms explaining risky needle use differ for drug user and sex partner networks in ways consistent with the functional nature of each type of network tie.

Adolescent↗
PubMed1991

The immune system as a neural network: a multi-epitope approach.

The term "neural network" has been applied to arrays of simple activation units linked by weighted connections. If the connections are modified according to a defined learning algorithm, such networks can be trained to store and retrieve patterned information. Memories are distributed throughout the network, allowing the network to recall complete patterns from incomplete input (pattern completion). The major biological application of neural network theory to date has been in the neurosciences, but the immune system may represent an alternative organ system in which to search for neural network architecture. Previous applications of parallel distributed processing to idiotype network theory have focused upon the recognition of individual epitopes. We argue here that this approach may be too restrictive, underestimating the power of neural network architecture. We propose that the network stores and retrieves large, complex patterns consisting of multiple epitopes separated in time and space. Such a network would be capable of perceiving an entire bacterium, and of storing the time course of a viral infection. While recognition of solitary epitopes occurs at the cellular level in this model, recognition of structures larger than the width of an antibody binding site takes place at the organ level, via network architecture integration of, i.e. individual epitope responses. The Oudin-Cazenave enigma, the sharing of idiotypic determinants by antibodies directed against distinct regions of the same antigen, suggests that some network level of integration of the individual clonal responses to large antigens does occur. The role of cytokines in prior neural network models of the immune system is unclear. We speculate that cytokines may influence the temperature of the network, such that changes in the cytokine milieu serve to "anneal" the network, allowing it to achieve the optimum steady-state in the shortest period of time.

Artificial Intelligence↗
PubMed2006

Cellular networks and the aging process.

The most important interactions between cellular molecules have a high affinity, are unique and specific, and require a network approach for a detailed description. After a brief introduction to cellular networks (protein--protein interaction networks, metabolic networks, gene regulatory networks, signalling networks and membrane--organelle networks) an overview is given on the network aspects of the theories on aging. The most important part of the review summarizes our knowledge on the aging of networks. The effects of aging on the general network models are described, as well as the initial findings on the effects of aging on the cellular networks. Finally we suggest a 'weak link theory of aging' linking the random damage of the network constituents to the overwhelming majority of the low affinity, transient interactions (weak links) in the cellular networks. We show that random damage of weak links may lead to an increase of noise and an increased vulnerability of cellular networks, and make a comparison between these predictions and the observed behaviour of the emergent properties of cellular networks in aged organisms.

Aging↗
PubMed2002

A duplication growth model of gene expression networks.

MOTIVATION: There has been considerable interest in developing computational techniques for inferring genetic regulatory networks from whole-genome expression profiles. When expression time series data sets are available, dynamic models can, in principle, be used to infer correlative relationships between gene expression levels, which may be causal. However, because of the range of detectable expression levels and the current quality of the data, the predictive nature of such inferred, quantitative models is questionable. Network models derived from simple rate laws offer an intermediate level analysis, going beyond simple statistical analysis, but falling short of a fully quantitative description. This work shows how such network models can be constructed and describes the global properties of the networks derived from such a model. These global properties are statistically robust and provide insights into the design of the underlying network. RESULTS: Several whole-genome expression time series data sets from yeast microarray experiments were analyzed using a Markov-modeling method (Dewey and Galas, FUNC: Integr. Genomics, 1, 269-278, 2001) to infer an approximation to the underlying genetic network. We found that the global statistical properties of all the resulting networks are similar. The overall structure of these biological networks is distinctly different from that of other recently studied networks such as the Internet or social networks. These biological networks show hierarchical, hub-like structures that have some properties similar to a class of graphs known as small world graphs. Small world networks exhibit local cliquishness while exhibiting strong global connectivity. In addition to the small world properties, the biological networks show a power law or scale free distribution of connectivities. An inverse power law, N(k) approximately k(-3/2), for the number of vertices (genes) with k connections was observed for three different data sets from yeast. We propose network growth models based on gene duplication events. Simulations of these models yield networks with the same combination of global graphical properties that we inferred from the expression data.

Algorithms↗
PubMed2003

The connectivity structure, giant strong component and centrality of metabolic networks.

MOTIVATION: Structural and functional analysis of genome-based large-scale metabolic networks is important for understanding the design principles and regulation of the metabolism at a system level. The metabolic network is conventionally considered to be highly integrated and very complex. A rational reduction of the metabolic network to its core structure and a deeper understanding of its functional modules are important. RESULTS: In this work, we show that the metabolites in a metabolic network are far from fully connected. A connectivity structure consisting of four major subsets of metabolites and reactions, i.e. a fully connected sub-network, a substrate subset, a product subset and an isolated subset is found to exist in metabolic networks of 65 fully sequenced organisms. The largest fully connected part of a metabolic network, called 'the giant strong component (GSC)', represents the most complicated part and the core of the network and has the feature of scale-free networks. The average path length of the whole network is primarily determined by that of the GSC. For most of the organisms, GSC normally contains less than one-third of the nodes of the network. This connectivity structure is very similar to the 'bow-tie' structure of World Wide Web. Our results indicate that the bow-tie structure may be common for large-scale directed networks. More importantly, the uncovered structure feature makes a structural and functional analysis of large-scale metabolic network more amenable. As shown in this work, comparing the closeness centrality of the nodes in the GSC can identify the most central metabolites of a metabolic network. To quantitatively characterize the overall connection structure of the GSC we introduced the term 'overall closeness centralization index (OCCI)'. OCCI correlates well with the average path length of the GSC and is a useful parameter for a system-level comparison of metabolic networks of different organisms. SUPPLEMENTARY INFORMATION: http://genome.gbf.de/bioinformatics/

Bacteria↗
PubMed2004

Efficient sampling algorithm for estimating subgraph concentrations and detecting network motifs.

SUMMARY: Biological and engineered networks have recently been shown to display network motifs: a small set of characteristic patterns that occur much more frequently than in randomized networks with the same degree sequence. Network motifs were demonstrated to play key information processing roles in biological regulation networks. Existing algorithms for detecting network motifs act by exhaustively enumerating all subgraphs with a given number of nodes in the network. The runtime of such algorithms increases strongly with network size. Here, we present a novel algorithm that allows estimation of subgraph concentrations and detection of network motifs at a runtime that is asymptotically independent of the network size. This algorithm is based on random sampling of subgraphs. Network motifs are detected with a surprisingly small number of samples in a wide variety of networks. Our method can be applied to estimate the concentrations of larger subgraphs in larger networks than was previously possible with exhaustive enumeration algorithms. We present results for high-order motifs in several biological networks and discuss their possible functions. AVAILABILITY: A software tool for estimating subgraph concentrations and detecting network motifs (mfinder 1.1) and further information is available at http://www.weizmann.ac.il/mcb/UriAlon/

Algorithms↗
PubMed1995

The permeability of fibrin network developed in human plasma.

Fibrin network permeability has an important role in thrombosis and inflammation since it influences the rate of transport of macromolecules through the network by convection. The conditions of polymerization of fibrin determine the network permeability and this has been attributed to variability in fibrin fibre thickness. Inconsistencies between values for fibrin fibre thickness derived from turbidity and permeability were examined. Networks were developed from human plasma by the addition of thrombin and network polymerization was modified pharmacologically. Dextran (MW 70,000) and poloxamer 188 both increased, and lauryl sulphate decreased, network permeability and network turbidity. Network fibre thickness was consistently higher when derived from permeability than from turbidity. Network permeability was significantly more susceptible to pharmacological manipulation by these agents than network turbidity. These inconsistencies were attributed to variation in the arrangement of the network fibres such as inhomogeneity of network fibre distribution and to fibre aggregation or alignment. Collectively these factors prohibit the derivation of fibrin fibre thickness from permeability. The dimensionless permeability (network permeability/(fibre radius)2) was used as an index of network fibre arrangement and found to be readily modified pharmacologically. Physiological and pharmacological regulation of fibrin network permeability may be predominantly mediated through modification of fibre arrangement and not through fibre thickness.

Biopolymers↗
PubMed2006

Topology of resultant networks shaped by evolutionary pressure.

Understanding the topology of complex systems abstracted to networks is important for unraveling their functional capabilities. Many such networks follow the small-world and scale-free regimes. Several models of artificially growing networks lead to this observed network topology. Most previously proposed models for growing networks, such as rich-get-richer and duplication-divergence, produce realistic network topologies but do not consider the effects of exogenous forces such as optimization for adaptation in shaping network topology. It is likely that such forces have shaped complex systems throughout their evolution. To develop further insights into possible mechanisms that shape networks, a model that uses several previously proposed network growth algorithms was developed to grow networks that adapt under exogenous stress. A decision tree problem was used to generate a complex Boolean function. Growing networks were required to adapt to correctly decode this function using an evolutionary selection process. Under this growth regimen all growing network models are similarly adaptable. The newly added nodes tend to cluster into pathways emanating from few inputs, regardless of the growth algorithm. Distribution of redundant pathways from inputs to the output follow a power-law function with a scaling exponent (approximately 1.3). Similar distribution of redundant pathways was observed from inputs in a cell signaling network and an air traffic control network. A flat distribution of redundant pathways from inputs was observed in growing networks that do not attempt to adapt. This analysis provides initial insights into distribution of pathways in naturally evolving complex systems that have defined input-output relationships.

Algorithms↗
PubMed2004

Sexual network analysis of a gonorrhoea outbreak.

OBJECTIVES: Sexual partnerships can be viewed as networks in order to study disease transmission. We examined the transmission of Neisseria gonorrhoeae in a localised outbreak in Alberta, Canada, using measures of network centrality to determine the association between risk of infection of network members and their position within the sexual network. We also compared risk in smaller disconnected components with a large network centred on a social venue. METHODS: During the investigation of the outbreak, epidemiological data were collected on gonorrhoea cases and their sexual contacts from STI surveillance records. In addition to traditional contact tracing information, subjects were interviewed about social venues they attended in the past year where casual sexual partnering may have occurred. Sexual networks were constructed by linking together named partners. Univariate comparisons of individual network member characteristics and algebraic measures of network centrality were completed. RESULTS: The sexual networks consisted of 182 individuals, of whom 107 were index cases with laboratory confirmed gonorrhoea and 75 partners of index cases. People who had significantly higher information centrality within each of their local networks were found to have patronised a popular motel bar in the main town in the region (p = 0.05). When the social interaction through the bar was considered, a large network of 89 individuals was constructed that joined all eight of the largest local networks. Moreover, several networks from different communities were linked by individuals who served as bridge populations as a result of their sexual partnering. CONCLUSION: Asking clients about particular social venues emphasised the importance of location in disease transmission. Network measures of centrality, particularly information centrality, allowed the identification of key individuals through whom infection could be channelled into local networks. Such individuals would be ideal targets for increased interventions.

Adolescent↗
PubMed1990

Cellular and synaptic mechanisms responsible for a long-lasting restructuring of the lobster pyloric network.

1. In the lobster Palinurus vulgaris a sensory input in the lateral posterolateral nerve (lpln) of the stomatogastric nervous system (STS) is able to turn on the cardiac sac (CS) network and to induce dramatic long-lasting alterations in the output of the pyloric network. This long-lasting alteration of pyloric network output consists primarily of changes in the activity of the two neurons that innervate the muscles of the cardiopyloric valve of the stomach, with the dilator neuron (the ventricular dilator, VD) transferring from the pyloric network to the CS network and the constrictor neuron (the inferior cardiac, IC) shifting to fire earlier in the pyloric pattern. 2. The inferior ventricular (IV) neurons of the CS network make complex multiaction synaptic connections onto several pyloric neurons in a related species, Panulirus interruptus. We show that many of the short-term alterations in pyloric activity observed during CS network bursts in Palinurus are due to similar IV neuron synaptic connections. However, the long-lasting effects of lpln stimulation on pyloric output are not due to this synaptic input, because 1) direct activation of the IV neurons does not induce long-lasting changes in pyloric activity and 2) pharmacologic disconnection of this synaptic input does not abolish lpln stimulation's long-lasting effects. Lpln stimulation therefore activates two different neuronal inputs to the pyloric network. 3. The transfer of the VD neuron from the pyloric to the CS network is the result of the concerted actions of these two inputs. Lpln stimulation turns on the CS network, and the IV neurons of the CS network excite the VD neuron and ensure it fires with the CS network. The second neuronal input (that not involving known CS network neurons) abolishes in a long-lasting fashion the VD neuron regenerative (plateau) properties, and thus suppresses the ability of the VD neuron to participate in the pyloric rhythmic pattern between CS network bursts. 4. Experimental manipulation of VD neuron activity can both mimic and reverse the effects of lpln stimulation on the IC neuron. The changes in IC neuron activity are therefore not due to direct lpln-activated synaptic input onto the IC neuron, but instead are indirect "network" effects arising from the changes in VD neuron activity.

Animals↗
PubMed2006

Design principles of molecular networks revealed by global comparisons and composite motifs.

BACKGROUND: Molecular networks are of current interest, particularly with the publication of many large-scale datasets. Previous analyses have focused on topologic structures of individual networks. RESULTS: Here, we present a global comparison of four basic molecular networks: regulatory, co-expression, interaction, and metabolic. In terms of overall topologic correlation--whether nearby proteins in one network are close in another--we find that the four are quite similar. However, focusing on the occurrence of local features, we introduce the concept of composite hubs, namely hubs shared by more than one network. We find that the three 'action' networks (metabolic, co-expression, and interaction) share the same scaffolding of hubs, whereas the regulatory network uses distinctly different regulator hubs. Finally, we examine the inter-relationship between the regulatory network and the three action networks, focusing on three composite motifs--triangles, trusses, and bridges--involving different degrees of regulation of gene pairs. Our analysis shows that interaction and co-expression networks have short-range relationships, with directly interacting and co-expressed proteins sharing regulators. However, the metabolic network contains many long-distance relationships: far-away enzymes in a pathway often have time-delayed expression relationships, which are well coordinated by bridges connecting their regulators. CONCLUSION: We demonstrate how basic molecular networks are distinct yet connected and well coordinated. Many of our conclusions can be mapped onto structured social networks, providing intuitive comparisons. In particular, the long-distance regulation in metabolic networks agrees with its counterpart in social networks (namely, assembly lines). Conversely, the segregation of regulator hubs from other hubs diverges from social intuitions (as managers often are centers of interactions).

Amino Acid Motifs↗
PubMed2006

Function approximation and documentation of sampling data using artificial neural networks.

Biodiversity studies in ecology often begin with the fitting and documentation of sampling data. This study is conducted to make function approximation on sampling data and to document the sampling information using artificial neural network algorithms, based on the invertebrate data sampled in the irrigated rice field. Three types of sampling data, i.e., the curve species richness vs. the sample size, the curve rarefaction, and the curve mean abundance of newly sampled species vs.the sample size, are fitted and documented using BP (Backpropagation) network and RBF (Radial Basis Function) network. As the comparisons, The Arrhenius model, and rarefaction model, and power function are tested for their ability to fit these data. The results show that the BP network and RBF network fit the data better than these models with smaller errors. BP network and RBF network can fit non-linear functions (sampling data) with specified accuracy and don't require mathematical assumptions. In addition to the interpolation, BP network is used to extrapolate the functions and the asymptote of the sampling data can be drawn. BP network cost a longer time to train the network and the results are always less stable compared to the RBF network. RBF network require more neurons to fit functions and generally it may not be used to extrapolate the functions. The mathematical function for sampling data can be exactly fitted using artificial neural network algorithms by adjusting the desired accuracy and maximum iterations. The total numbers of functional species of invertebrates in the tropical irrigated rice field are extrapolated as 140 to 149 using trained BP network, which are similar to the observed richness.

Algorithms↗
PubMedDate not supplied

A self-organising network that grows when required.

The ability to grow extra nodes is a potentially useful facility for a self-organising neural network. A network that can add nodes into its map space can approximate the input space more accurately, and often more parsimoniously, than a network with predefined structure and size, such as the Self-Organising Map. In addition, a growing network can deal with dynamic input distributions. Most of the growing networks that have been proposed in the literature add new nodes to support the node that has accumulated the highest error during previous iterations or to support topological structures. This usually means that new nodes are added only when the number of iterations is an integer multiple of some pre-defined constant, A. This paper suggests a way in which the learning algorithm can add nodes whenever the network in its current state does not sufficiently match the input. In this way the network grows very quickly when new data is presented, but stops growing once the network has matched the data. This is particularly important when we consider dynamic data sets, where the distribution of inputs can change to a new regime after some time. We also demonstrate the preservation of neighbourhood relations in the data by the network. The new network is compared to an existing growing network, the Growing Neural Gas (GNG), on a artificial dataset, showing how the network deals with a change in input distribution after some time. Finally, the new network is applied to several novelty detection tasks and is compared with both the GNG and an unsupervised form of the Reduced Coulomb Energy network on a robotic inspection task and with a Support Vector Machine on two benchmark novelty detection tasks.

Algorithms↗
PubMed2006

A methodology for the structural and functional analysis of signaling and regulatory networks.

BACKGROUND: Structural analysis of cellular interaction networks contributes to a deeper understanding of network-wide interdependencies, causal relationships, and basic functional capabilities. While the structural analysis of metabolic networks is a well-established field, similar methodologies have been scarcely developed and applied to signaling and regulatory networks. RESULTS: We propose formalisms and methods, relying on adapted and partially newly introduced approaches, which facilitate a structural analysis of signaling and regulatory networks with focus on functional aspects. We use two different formalisms to represent and analyze interaction networks: interaction graphs and (logical) interaction hypergraphs. We show that, in interaction graphs, the determination of feedback cycles and of all the signaling paths between any pair of species is equivalent to the computation of elementary modes known from metabolic networks. Knowledge on the set of signaling paths and feedback loops facilitates the computation of intervention strategies and the classification of compounds into activators, inhibitors, ambivalent factors, and non-affecting factors with respect to a certain species. In some cases, qualitative effects induced by perturbations can be unambiguously predicted from the network scheme. Interaction graphs however, are not able to capture AND relationships which do frequently occur in interaction networks. The consequent logical concatenation of all the arcs pointing into a species leads to Boolean networks. For a Boolean representation of cellular interaction networks we propose a formalism based on logical (or signed) interaction hypergraphs, which facilitates in particular a logical steady state analysis (LSSA). LSSA enables studies on the logical processing of signals and the identification of optimal intervention points (targets) in cellular networks. LSSA also reveals network regions whose parametrization and initial states are crucial for the dynamic behavior. We have implemented these methods in our software tool CellNetAnalyzer (successor of FluxAnalyzer) and illustrate their applicability using a logical model of T-Cell receptor signaling providing non-intuitive results regarding feedback loops, essential elements, and (logical) signal processing upon different stimuli. CONCLUSION: The methods and formalisms we propose herein are another step towards the comprehensive functional analysis of cellular interaction networks. Their potential, shown on a realistic T-cell signaling model, makes them a promising tool.

Animals↗