Yung-Keun Kwon

A Hybrid Neurogenetic Approach for Stock Forecasting

In this paper, we propose a hybrid neurogenetic system for stock trading. A recurrent neural network (NN) having one hidden layer is used for the prediction model. The input features are generated from a number of technical indicators being used by financial experts. The genetic algorithm (GA) optimizes the NNs weights under a 2-D encoding and crossover. We devised a context-based ensemble method of NNs which dynamically changes on the basis of the test days context. To reduce the time in processing mass data, we parallelized the GA on a Linux cluster system using message passing interface. We tested the proposed method with 36 companies in NYSE and NASDAQ for 13 years from 1992 to 2004. The neurogenetic hybrid showed notable improvement on the average over the buy-and-hold strategy and the context-based ensemble further improved the results. We also observed that some companies were more predictable than others, which implies that the proposed neurogenetic hybrid can be used for financial portfolio construction

Quantitative analysis of robustness and fragility in biological networks based on feedback dynamics

MOTIVATION It has been widely reported that biological networks are robust against perturbations such as mutations. On the contrary, it has also been known that biological networks are often fragile against unexpected mutations. There is a growing interest in these intriguing observations and the underlying design principle that causes such robust but fragile characteristics of biological networks. For relatively small networks, a feedback loop has been considered as an important motif for realizing the robustness. It is still, however, not clear how a number of coupled feedback loops actually affect the robustness of large complex biological networks. In particular, the relationship between fragility and feedback loops has not yet been investigated till now. RESULTS Through extensive computational experiments, we found that networks with a larger number of positive feedback loops and a smaller number of negative feedback loops are likely to be more robust against perturbations. Moreover, we found that the nodes of a robust network subject to perturbations are mostly involved with a smaller number of feedback loops compared with the other nodes not usually subject to perturbations. This topological characteristic eventually makes the robust network fragile against unexpected mutations at the nodes not previously exposed to perturbations.

Coherent coupling of feedback loops

MOTIVATION It is widely accepted that cell signaling networks have been evolved to be robust against perturbations. To investigate the topological characteristics resulting in such robustness, we have examined large-scale signaling networks and found that a number of feedback loops are present mostly in coupled structures. In particular, the coupling was made in a coherent way implying that same types of feedback loops are interlinked together. RESULTS We have investigated the role of such coherently coupled feedback loops through extensive Boolean network simulations and found that a high proportion of coherent couplings can enhance the robustness of a network against its state perturbations. Moreover, we found that the robustness achieved by coherently coupled feedback loops can be kept evolutionarily stable. All these results imply that the coherent coupling of feedback loops might be a design principle of cell signaling networks devised to achieve the robustness.

GPEC: A Cytoscape plug-in for random walk-based gene prioritization and biomedical evidence collection

Finding genes associated with a disease is an important issue in the biomedical area and many gene prioritization methods have been proposed for this goal. Among these, network-based approaches are recently proposed and outperformed functional annotation-based ones. Here, we introduce a novel Cytoscape plug-in, GPEC, to help identify putative genes likely to be associated with specific diseases or pathways. In the plug-in, gene prioritization is performed through a random walk with restart algorithm, a state-of-the art network-based method, along with a gene/protein relationship network. The plug-in also allows users efficiently collect biomedical evidence for highly ranked candidate genes. A set of known genes, candidate genes and a gene/protein relationship network can be provided in a flexible way.

Analysis of feedback loops and robustness in network evolution based on Boolean models

BackgroundMany biological networks such as protein-protein interaction networks, signaling networks, and metabolic networks have topological characteristics of a scale-free degree distribution. Preferential attachment has been considered as the most plausible evolutionary growth model to explain this topological property. Although various studies have been undertaken to investigate the structural characteristics of a network obtained using this growth model, its dynamical characteristics have received relatively less attention.ResultsIn this paper, we focus on the robustness of a network that is acquired during its evolutionary process. Through simulations using Boolean network models, we found that preferential attachment increases the number of coupled feedback loops in the course of network evolution. Whereas, if networks evolve to have more coupled feedback loops rather than following preferential attachment, the resulting networks are more robust than those obtained through preferential attachment, although both of them have similar degree distributions.ConclusionThe presented analysis demonstrates that coupled feedback loops may play an important role in network evolution to acquire robustness. The result also provides a hint as to why various biological networks have evolved to contain a number of coupled feedback loops.

Stock prediction based on financial correlation

In this paper, we propose a neuro-genetic stock prediction system based on financial correlation between companies. A number of input variables are produced from the relatively highly correlated companies. The genetic algorithm selects a set of informative input features among them for a recurrent neural network. It showed notable improvement over not only the buy-and-hold strategy but also the recurrent neural network using only the input variables from the target company.

Neighbor-favoring weight reinforcement to improve random walk-based disease gene prioritization

BACKGROUND Finding candidate genes associated with a disease is an important issue in biomedical research. Recently, many network-based methods have been proposed that implicitly utilize the modularity principle, which states that genes causing the same or similar diseases tend to form physical or functional modules in gene/protein relationship networks. Of these methods, the random walk with restart (RWR) algorithm is considered to be a state-of-the-art approach, but the modularity principle has not been fully considered in traditional RWR approaches. Therefore, we propose a novel method called ORIENT (neighbor-favoring weight reinforcement) to improve the performance of RWR through proper intensification of the weights of interactions close to the known disease genes. RESULTS Through extensive simulations over hundreds of diseases, we observed that our approach performs better than the traditional RWR algorithm. In particular, our method worked best when the weights of interactions involving only the nearest neighbor genes of the disease genes were intensified. Interestingly, the performance of our approach was negatively related to the probability with which the random walk will restart, whereas the performance of RWR without the weight-reinforcement was positively related in dense gene/protein relationship networks. We further found that the density of the disease gene-projected sub-graph and the number of paths between the disease genes in a gene/protein relationship network may be explanatory variables for the RWR performance. Finally, a comparison with other well-known gene prioritization tools including Endeavour, ToppGene, and BioGraph, revealed that our approach shows significantly better performance. CONCLUSION Taken together, these findings provide insight to efficiently guide RWR in disease gene prioritization.

Reduction of complex signaling networks to a representative kernel.

An algorithmic approach enables the simplification of complex signaling networks and identifies potential therapeutic targets. Reducing Complexity The large and complex nature of the biochemical regulatory networks that govern cell behavior provides a major challenge to the systematic analysis of cell signaling. However, most processes that reduce network complexity fail to reproduce the dynamic properties of the original network. Kim et al. describe an algorithmic approach to network reduction and simplification that preserves the dynamics of the network. They applied their approach to several networks in species from bacteria to humans, producing simplified networks called “kernels.” Examination of the genes represented by the kernel nodes provided insight into the evolution of these core network genes. Furthermore, the genes represented by the kernel nodes were enriched in disease-associated genes and drug targets, suggesting that this type of analysis may be therapeutically beneficial. The network of biomolecular interactions that occurs within cells is large and complex. When such a network is analyzed, it can be helpful to reduce the complexity of the network to a “kernel” that maintains the essential regulatory functions for the output under consideration. We developed an algorithm to identify such a kernel and showed that the resultant kernel preserves the network dynamics. Using an integrated network of all of the human signaling pathways retrieved from the KEGG (Kyoto Encyclopedia of Genes and Genomes) database, we identified this network’s kernel and compared the properties of the kernel to those of the original network. We found that the percentage of essential genes to the genes encoding nodes outside of the kernel was about 10%, whereas ~32% of the genes encoding nodes within the kernel were essential. In addition, we found that 95% of the kernel nodes corresponded to Mendelian disease genes and that 93% of synthetic lethal pairs associated with the network were contained in the kernel. Genes corresponding to nodes in the kernel had low evolutionary rates, were ubiquitously expressed in various tissues, and were well conserved between species. Furthermore, kernel genes included many drug targets, suggesting that other kernel nodes may be potential drug targets. Owing to the simplification of the entire network, the efficient modeling of a large-scale signaling network and an understanding of the core structure within a complex framework become possible.

A 35-60 GHz single-pole double-throw(SPDT) switching circuit using direct contact MEMS switches and double resonance technique

In this paper, a single-pole double-throw(SPDT) switching circuit was demonstrated using direct contact MEMS switches and double resonance technique for Q-band and V-band applications. The size of the fabricated SPDT switching circuit is about 1 mm/spl times/2 mm. The direct contact MEMS switches are formed on the CPW transmission lines and actuated with electrostatic force. The fabricated single-pole single-throw(SPST) MEMS switch shows the insertion loss of 0.34 dB and the isolation of 15.5 dB at 50 GHz, respectively. The responses of the fabricated SPDT switching circuit were measured with the frequencies from 0 to 100 GHz. The insertion loss is below 1 dB and isolation is better than 19 dB from 35 GHz to 60 GHz. At the center frequency of 47 GHz, the insertion loss is measured 0.45 dB and isolation is 22 dB. The actuation voltage of the switch is 35V.

Investigations into the relationship between feedback loops and functional importance of a signal transduction network based on Boolean network modeling

BackgroundA number of studies on biological networks have been carried out to unravel the topological characteristics that can explain the functional importance of network nodes. For instance, connectivity, clustering coefficient, and shortest path length were previously proposed for this purpose. However, there is still a pressing need to investigate another topological measure that can better describe the functional importance of network nodes. In this respect, we considered a feedback loop which is ubiquitously found in various biological networks.ResultsWe discovered that the number of feedback loops (NuFBL) is a crucial measure for evaluating the importance of a network node and verified this through a signal transduction network in the hippocampal CA1 neuron of mice as well as through generalized biological network models represented by Boolean networks. In particular, we observed that the proteins with a larger NuFBL are more likely to be essential and to evolve slowly in the hippocampal CA1 neuronal signal transduction network. Then, from extensive simulations based on the Boolean network models, we proved that a network node with the larger NuFBL is likely to be more important as the mutations of the initial state or the update rule of such a node made the network converge to a different attractor. These results led us to infer that such a strong positive correlation between the NuFBL and the importance of a network node might be an intrinsic principle of biological networks in view of network dynamics.ConclusionThe presented analysis on topological characteristics of biological networks showed that the number of feedback loops is positively correlated with the functional importance of network nodes. This result also suggests the existence of unknown feedback loops around functionally important nodes in biological networks.