Haluk Bingol

Multiway analysis of epilepsy tensors

MOTIVATION The success or failure of an epilepsy surgery depends greatly on the localization of epileptic focus (origin of a seizure). We address the problem of identification of a seizure origin through an analysis of ictal electroencephalogram (EEG), which is proven to be an effective standard in epileptic focus localization. SUMMARY With a goal of developing an automated and robust way of visual analysis of large amounts of EEG data, we propose a novel approach based on multiway models to study epilepsy seizure structure. Our contributions are 3-fold. First, we construct an Epilepsy Tensor with three modes, i.e. time samples, scales and electrodes, through wavelet analysis of multi-channel ictal EEG. Second, we demonstrate that multiway analysis techniques, in particular parallel factor analysis (PARAFAC), provide promising results in modeling the complex structure of an epilepsy seizure, localizing a seizure origin and extracting artifacts. Third, we introduce an approach for removing artifacts using multilinear subspace analysis and discuss its merits and drawbacks. RESULTS Ictal EEG analysis of 10 seizures from 7 patients are included in this study. Our results for 8 seizures match with clinical observations in terms of seizure origin and extracted artifacts. On the other hand, for 2 of the seizures, seizure localization is not achieved using an initial trial of PARAFAC modeling. In these cases, first, we apply an artifact removal method and subsequently apply the PARAFAC model on the epilepsy tensor from which potential artifacts have been removed. This method successfully identifies the seizure origin in both cases.

Seizure Recognition on Epilepsy Feature Tensor

With a goal of automating visual analysis of electroencephalogram (EEG) data and assessing the performance of various features in seizure recognition, we introduce a mathematical model capable of recognizing patient-specific epileptic seizures with high accuracy. We represent multi-channel scalp EEG using a set of features. These features expected to have distinct trends during seizure and non-seizure periods include features from both time and frequency domains. The contributions of this paper are threefold. First, we rearrange multi-channel EEG signals as a third-order tensor called an Epilepsy Feature Tensor with modes: time epochs, features and electrodes. Second, we model the Epilepsy Feature Tensor using a multilinear regression model, i.e., Multilinear Partial Least Squares regression, which is the generalization of Partial Least Squares (PLS) regression to higher-order datasets. This two-step approach facilitates EEG data analysis from multiple electrodes represented by several features from different domains. Third, we identify which features are more significant for seizure recognition. Our results based on the analysis of 19 seizures from 5 epileptic patients demonstrate that multiway analysis of an Epilepsy Feature Tensor can detect (patient-specific) seizures with classification accuracy ranging between 77-96%.

A formal treatment of generalized preferential attachment and its empirical validation

Generalized preferential attachment is defined as the tendency of a vertex to acquire new links in the future with respect to a particular vertex property. Understanding which properties influence the link acquisition tendency (LAT) gives us a predictive power to estimate the future growth of network and insight about the actual dynamics governing the complex networks. In this study, we explore the effect of age and degree on LAT by analyzing data collected from a new complex-network growth dataset. We found that LAT and degree of a vertex are linearly correlated in accordance with previous studies. Interestingly, the relation between LAT and age of a vertex is found to be in conflict with the known models of network growth. We identified three different periods in the networks lifetime where the relation between age and LAT is strongly positive, almost stationary and negative correspondingly.

Social Network of Co-occurrence in News Articles

Networks describe various complex natural systems including social systems. Recent studies have shown that these networks share some common properties. While studying complex systems, data collection phase is difficult for social networks compared to other networks such as the WWW, Internet, protein or linguistic networks. Many interesting social networks such as movie actors’ collaboration, scientific collaboration and sexual contacts have been studied in the literature. It has been shown that they have small-world and power-law degree distribution properties. In this paper, we investigate an interesting social network of co-occurrence in news articles with respect to small-world and power-law degree distribution properties. 3000 news articles selected from Reuters-21578 corpus, which consists of news articles that appeared in the Reuters newswire in 1987 are used as the data set. Results reveal that like the previously studied social networks the social network of co-occurrence in news articles also possesses the small-world and power-law degree distribution properties.

Graph-theoretic analysis of epileptic seizures on scalp EEG recordings

This work presents a novel modeling of neuronal activity of the brain by capturing the synchronization of EEG signals along the scalp. The pair-wise correspondence between electrodes recording EEG signals are used to establish edges between these electrodes which then become the nodes of a synchronization graph. As EEG signals are recorded over time, we discretize the time axis into overlapping epochs, and build a series of time-evolving synchronization graphs for each epoch and for each traditional frequency band. We show that graph theory provides a rich set of graph features that can be used for mining and learning from the EEG signals to determine temporal and spatial localization of epileptic seizures. We present several techniques to capture the pair-wise synchronization and apply unsupervised learning algorithms, such as k-means clustering and multiway modeling of third-order tensors, to analyze the labeled clinical data in the feature domain to detect the onset and origin location of the seizure. We use k-means clustering on two-way feature matrices for detection of seizures, and Tucker3 tensor decomposition for localization of seizures. We conduct an extensive parametric search to determine the best configuration of the model parameters including epoch length, synchronization metrics, and frequency bands, to achieve the highest accuracy. Our results are promising: we are able to detect the onset of seizure with an accuracy of 88.24%, and localize the onset of the seizure with an accuracy of 76.47%.

Community detection using preference networks

Abstract Community detection is the task of identifying clusters or groups of nodes in a network where nodes within the same group are more connected with each other than with nodes in different groups. It has practical uses in identifying similar functions or roles of nodes in many biological, social and computer networks. With the availability of very large networks in recent years, performance and scalability of community detection algorithms become crucial, i.e. if time complexity of an algorithm is high, it cannot run on large networks. In this paper, we propose a new community detection algorithm, which has a local approach and is able to run on large networks. It has a simple and effective method; given a network, algorithm constructs a preference network of nodes where each node has a single outgoing edge showing its preferred node to be in the same community with. In such a preference network, each connected component is a community. Selection of the preferred node is performed using similarity based metrics of nodes. We use two alternatives for this purpose which can be calculated in 1-neighborhood of nodes, i.e. number of common neighbors of selector node and its neighbors and, the spread capability of neighbors around the selector node which is calculated by the gossip algorithm of Lind et.al. Our algorithm is tested on both computer generated LFR networks and real-life networks with ground-truth community structure. It can identify communities accurately in a fast way. It is local, scalable and suitable for distributed execution on large networks.

Attention competition with advertisement.

In the new digital age, information is available in large quantities. Since information consumes primarily the attention of its recipients, the scarcity of attention is becoming the main limiting factor. In this study, we investigate the impact of advertisement pressure on a cultural market where consumers have a limited attention capacity. A model of competition for attention is developed and investigated analytically and by simulation. Advertisement is found to be much more effective when the attention capacity of agents is extremely scarce. We have observed that the market share of the advertised item improves if dummy items are introduced to the market while the strength of the advertisement is kept constant.

GOSSIP ON WEIGHTED NETWORKS

In this work, we analyze gossip spreading on weighted networks. We try to define a new metric to classify weighted complex networks using our model. The model proposed here is based on the gossip spreading model introduced by Lind et al. on unweighted networks. The new metric is based on gossip spreading activity in the network, which is correlated with both topology and relative edge weights in the network. The model gives more insight about the weight distribution and correlation of topology with edge weights in a network. It also measures how suitable a weighted network is for gossip spreading. We analyze gossip spreading on real weighted networks of human interactions. Six co-occurrence and seven social pattern networks are investigated. Gossip propagation is found to be a good parameter to distinguish co-occurrence and social pattern networks. As a comparison some miscellaneous networks of comparable sizes and computer generated networks based on ER, BA and WS models are also investigated. They are found to be quite different from the human interaction networks.

Context sensitive article ranking with citation context analysis

It is hard to detect important articles in a specific context. Information retrieval techniques based on full text search can be inaccurate to identify main topics and they are not able to provide an indication about the importance of the article. Generating a citation network is a good way to find most popular articles but this approach is not context aware. The text around a citation mark is generally a good summary of the referred article. So citation context analysis presents an opportunity to use the wisdom of crowd for detecting important articles in a context sensitive way. In this work, we analyze citation contexts to rank articles properly for a given topic. The model proposed uses citation contexts in order to create a directed and edge-labeled citation network based on the target topic. Then we apply common ranking algorithms in order to find important articles in this newly created network. We showed that this method successfully detects a good subset of most prominent articles in a given topic. The biggest contribution of this approach is that we are able to identify important articles for a given search term even though these articles do not contain this search term. This technique can be used in other linked documents including web pages, legal documents, and patents as well as scientific papers.

Complex Networks in Different Languages: A Study of an Emergent Multilingual Encyclopedia

There is an increasing interest to the study of complex networks in an interdisciplinary way. Language, as a complex network, has been a part of this study due to its importance in human life. Moreover, the Internet has also been at the center of this study by making access to large amounts of information possible. With these ideas in mind, this work aims to evaluate conceptual networks in different languages with the data from a large and open source of information in the Internet, namely Wikipedia. As an evolving multilingual encyclopedia that can be edited by any Internet user, Wikipedia is a good example of an emergent complex system. In this paper, different from previous work on conceptual networks which usually concentrated on single languages, we concentrate on possible ways to compare the usages of different languages and possibly the underlying cultures. This also involves the analysis of local network properties around certain coneepts in different languages. For an initial evaluation, the concept “family” is used to compare the English and German Wikipedias. Although, the work is currently at the beginning, the results are promising.