I go by the name of “Mon” on the Internet when I need to take part in real-life activities. I also give this name to people to use as my nickname in the real life.
I am working on network science as a C.S. Ph.D. candidate. I was a computational maths undergraduate. I also took a Master’s under Computer Engineering but graduated with a thesis on network analysis. I thought I would be working on bioinformatics so I decided to change my major from life science to maths at the end of the freshman year, but I end up working with graphs.
I am aiming at cultivating an expertise of using network science for effective data representation which facilitates knowledge discovery. I strongly believe that good representation accomplishes 90% of the problem solving.
This is my tech-blog where I write about stable and reproducible stuff: some tools, some methods, and some designs.
Topics of this blog
Please refer to the “Category” page for a quick look.
My Philosophy
In general, I’m a devoted structualist. I like to strip things down to the bones – however to the contrary of popular beliefs that does not mean I’m such a reductionist, for experience has taught me hard lessons about the power of redundancy, and it is general believed that the knowledge is stored and retrived in a clustering manner, where the “neighbourhood” is the building-block for these clusters.
Graphs and networks
Graphs/networks are my go-to representation for most problems. When you throw relationships and findings into graph it seems to lazily store up the information, but the unstructured is only an illusion. I will write about how to extract the sequential and directional information from the only implicitly structured data so that it could be understood and used.
Neighbourhood: The “zero” hypothesis
When I say that I like the graph representation, I actually mean I like the “neighbourhood” representation. The “neighbourhood” paradigm has manifested itself across domains like social networks, NLP, recommendation, and cognitive science. In a network the local node information is permeated and shared by a group of nodes. This process is facilitated by the overlapping of neighbourhoods. As a structuralist, I’m interested in the utility/behavioral change in the network when I enhance/reduce the structural connections.
Completeness for graph data representation
Currently, I use graph for data discovery tasks in research, so I’m also interested in some computational problems related to the graph. For example, the size and density of the graph enabling a “full” representation of the data – the subgraph and graph sampling problems.