This is the personal homepage of Lazaros Gallos.
I am an Associate Director and Research Professor at DIMACS at Rutgers University, where I am also directing the long-running DIMACS REU summer program. DIMACS fosters research and educational programs on topics that lie at the interface of discrete mathematics and theoretical computer science.
I am a member of the Editorial Boards at Scientific Reports, at PLOS ONE, and at Entropy. I have also been an Associate Editor for the highly selective APS journal Physical Review X.
In the past, I was a Research Associate at the Department of Ecology of the Rutgers University, working with Prof. Nina Fefferman. Before that, I was working with Prof. Hernan Makse at the City College of New York.
I received my PhD in Computational Physics from the Physics Dept. in Univ. of Thessaloniki (advisor Prof. Panos Argyrakis).
My research interests are broad and cover a lot of inter-disciplinary ground. For the last few years I have been working on complex networks science. Currently, I work on (a) understanding fundamental mechanisms behind online social interactions, (b) epidemics spreading, (c) modules organization in the brain, and (d) applications of Machine Learning in Network Science.
In the majority of network growth models, new nodes attach themselves to one or more existing nodes. Our propinquity model is based on the assumption that the selection of the second connection can be correlated to the position of the first connection. By varying the strength of this correlation, we can reproduce a variety of network structures and we can additionally tune the local density of the network.
The destruction of networks is well-understood, mainly through percolation concepts. The status of a network after destruction, though, is largely unexplored. We demonstrated that the isolated clusters remain quite close to each other and it is relatively easy to restore connectivity based on decisions made strictly on local information and minimal cost.
There are many practical applications where we want to compare two structures, such as complex networks. We proposed a method where we sample two networks at a given scale and compare the link density distribution at this scale. The method works very well, and can be used to identify effective classifiers.To compare two networks, you can download the source code ntangle.c. Check also the README file.
What are our motivations in choosing our online friends? In our recent paper we study high-quality data and show how we can estimate the influence of different social drivers. In simpler words, how probable is it that we reply to friend requests or how often do we connect to popular people? Although we may not realize this, our connections evolve and we may find ourselves in an environment different from our choices.
We recently finished a work on how obesity, health indicators, economic indicators, etc spread geographically. For obesity, we find strong spatial correlations that extend up to 1000 miles. This suggests that the obesity epidemic is a collective phenomenon.
Brain presents a conundrum: we need isolated modules to process information, but this should all result in one unified picture. Our network analysis of fMRI data suggests that brain modules are fractal (large-world) but they are connected through lower-strength connections, that result in a global small-world network. This is a concept similar with Granovetter's 'strength of the weak ties' in social networks.
You want to spread a message in a social network. What would be your choice of the starting point for more efficient spreading? Surprisingly, it is not always the most connected individual; location can be more important than the number of connections.