Searcharxiv⌕ Search

arXiv subjects

Mukkai Krishnamoorthy

Publications and source records attributed to Mukkai Krishnamoorthy.

7 recordsLinked to original sources
arXiv2017

Interactive Complexity: Software Metrics from an Ecosystem Perspective

With even the most trivial of applications now being written on top of millions of lines code of libraries, API's, and programming languages, much of the complexity that used to exist when designing software has been abstracted away to allow programmers to focus on primarily business logic. With each application relying so heavily on the ecosystem it was designed to run in, whether that is limited to a local system or includes dependencies on machines connected by networks, measuring the complexity of these systems can no longer be done simply by observing the code internal to the application; we also need to account for its external interactions. This is especially important when considering issues of security, which becomes more vital as our healthcare, financial, and automobiles rely on complicated software systems. We propose Interactive Complexity, which provide a quantitative measure of how intertwined parts of the system are. Some of the most well-known software complexity metrics out there are the metrics in the CK-metric suite; these metrics are designed for use in measuring object oriented systems, but we believe they can be adapted to help measure the interaction of software systems. Our experimental results show strong correlations between the number of bugs fixed in a release and the value of some of these metrics in systems of sufficient scale.

cs.SE↗
arXiv2014

Case study to approaches to finding patterns in citation networks

Analysis of a dataset including a network of LED patents and their metadata is carried out using several methods in order to answer questions about the domain. We are interested in finding the relationship between the metadata and the network structure; for example, are central patents in the network produced by larger or smaller companies? We begin by exploring the structure of the network without any metadata, applying known techniques in citation analysis and a simple clustering scheme. These techinques are then combined with metadata analysis to draw preliminary conclusions about the dataset.

cs.DL↗
arXiv2012

Exploration of Network Scaling: Variations on Optimal Channel Networks

Metabolic allometry, a common pattern in nature, is a close-to-3/4-power scaling law between metabolic rate and body mass in organisms, across and within species. An analogous relationship between metabolic rate and water volume in river networks has also been observed. Optimal Channel Networks (OCNs), at local optima, accurately model many scaling properties of river systems, including metabolic allometry. OCNs are embedded in two-dimensional space; this work extends the model to three dimensions. In this paper we compare characteristics of 3d OCNs with 2d OCNs and with organic metabolic networks, studying the scaling behaviors of area, length, volume, and energy. In addition, we take a preliminary look at comparing Steiner trees with OCNs. We find that the three-dimensional OCN has predictable characteristics analogous to those of the two-dimensional version, as well as scaling properties similar to metabolic networks in biological organisms.

math.OC↗
arXiv2012

Temporal Analysis of Literary and Programming Prose

Literary works reference a variety of globally shared themes including well-known people, events, and time periods. It is particularly interesting to locate patterns that are either invariant across time or exhibit a characteristic change across time, as they could imply something important about society that those works record. This paper suggests the use of Google n-gram viewer as a fast prototyping method for examining time-based properties over a rich sample of literary prose. Using this method, we find that some repeating periods of time, like Sunday, are referenced disproportionally, allowing us to pose questions such as why a day like Thursday is so unpopular. Furthermore, by treating software as a work of prose, we can apply a similar analysis to open-source software repositories and explore time-based relations in commit logs. Doing a simple statistical analysis on a few temporal keywords in the log records, we reinforce and weaken a few beliefs on how college students approach open source software. Finally, we help readers working on their own temporal analysis by comparing the fundamental differences between literary works and code repositories, and suggest blog or wiki as recently-emerging works.

cs.SE↗
arXiv2010

Algorithmic Detection of Computer Generated Text

Computer generated academic papers have been used to expose a lack of thorough human review at several computer science conferences. We assess the problem of classifying such documents. After identifying and evaluating several quantifiable features of academic papers, we apply methods from machine learning to build a binary classifier. In tests with two hundred papers, the resulting classifier correctly labeled papers either as human written or as computer generated with no false classifications of computer generated papers as human and a 2% false classification rate for human papers as computer generated. We believe generalizations of these features are applicable to similar classification problems. While most current text-based spam detection techniques focus on the keyword-based classification of email messages, a new generation of unsolicited computer-generated advertisements masquerade as legitimate postings in online groups, message boards and social news sites. Our results show that taking the formatting and contextual clues offered by these environments into account may be of central importance when selecting features with which to identify such unwanted postings.

stat.ML↗
arXiv2010

A Hybrid Model for Disease Spread and an Application to the SARS Pandemic

Pandemics can cause immense disruption and damage to communities and societies. Thus far, modeling of pandemics has focused on either large-scale difference equation models like the SIR and the SEIR models, or detailed micro-level simulations, which are harder to apply at a global scale. This paper introduces a hybrid model for pandemics considering both global and local spread of infections. We hypothesize that the spread of an infectious disease between regions is significantly influenced by global traffic patterns and the spread within a region is influenced by local conditions. Thus we model the spread of pandemics considering the connections between regions for the global spread of infection and population density based on the SEIR model for the local spread of infection. We validate our hybrid model by carrying out a simulation study for the spread of SARS pandemic of 2002-2003 using available data on population, population density, and traffic networks between different regions. While it is well-known that international relationships and global traffic patterns significantly influence the spread of pandemics, our results show that integrating these factors into relatively simple models can greatly improve the results of modeling disease spread.

cs.MA↗