Computer Science Research Seminar
The Computer Science Research Seminar is an exciting forum intended to provide faculty and students with an opportunity to share their latest research results, to exchange ideas, and to seek new research directions and collaborations.
The Rise of Functional Languages in the Cloud
Tuesday, April 5
Speaker: Louie Bacaj '08
Cloud computing has had a large effect in our industry in recent years, most startups today use the public cloud as the backbone and infrastructure of their respective firms. Due to this phenomena much of the way we build software has changed, specifically the tools we use to build for the cloud tend not to be the same tools we have traditionally used. Functional languages which were, for quite some time, relegated to academic circles are seeing a sort of renaissance and gaining mainstream support with the rise of cloud computing. Bacaj will dive into what it means to build on the public cloud and why you should learn a few functional concepts to help compliment your skill-set coming out of school. Ultimately this will help you in your search to get into our industry and some of my tips around functional concepts might even help you get ahead in your career.
New Techniques and Results in Authorship Attribution of Thomas Paine Works
Tuesday, April 5
Speakers: Lubomir Ivanov, Ph.D., Computer Science Department; Gary Berton, Institute for Thomas Paine Studies; Sean Campbell '16
We will present some new results and insights based on the use of alliteration for authorship attribution. We will also present three applications of our methodology: a de-attribution of a work previously assumed to have been authored by Thomas Paine's ("African Slavery in America"), an example from a still unknown author, and a newly discovered anti-slavery article by Thomas Paine.
Clojure, Made Simple
Thursday, March 31
Speaker: R. Hickey
In the eight years following its initial release, Clojure has become a popular alternative language on the JVM, seeing production use at financial firms, major retailers, analytics companies, and startups large and small. It has done so while remaining decidedly alternative―eschewing object orientation for functional programming, C-derived syntax for code-as-data, static typing for dynamic typing, REPL-driven development,and so on. Underpinning these differences is a commitment to the principle that we should be building our systems out of fundamentally simpler materials. This session looks at what makes Clojure different and why.
Mediated Semi-Quantum Cryptography
Tuesday, March 15, Noon
Room M132, Murphy Center
Speaker: Walter Krawec, Ph.D., Computer Science Department
Quantum key distribution (QKD) protocols permit two users to establish a secret key, secure against an all-powerful adversary – something impossible to do using only classical means. However, to operate, both users need the capability of manipulating quantum resources (e.g., preparing and measuring quantum bits - qubits - in a variety of bases). A semi-quantum key distribution (SQKD) protocol asks the question: "Can this be done if one of the two users is limited and, in a way, "classical" in nature?" (i.e., one of the two users cannot fully manipulate qubits while the other user can). The answer is yes. We recently asked the question: "Can both users be limited and classical in nature?" We proved the answer is still positive if they rely on an untrusted quantum server (someone who prepares and measures qubits but whom neither A nor B trust – in fact the server may be the all-powerful adversary). We will discuss our latest work in this area. All required concepts and notation will be introduced including: qubits, measurements and entanglement.
Using Networks to Study Genetic Factors Associated with Disease
Tuesday, March 8, 2016
Speakers: Aaron Kershenbaum, Computer Science Department; Robert Schiaffino, Ph.D., Computer Science Department; Keitha Murray, Computer Science Department
A wealth of genetic information is becoming available, both global information about the human genome and clinical information about specific people. There is great interest in mining this information to develop diagnostics and interventions aimed at specific people and diseases. This field is referred to as personalized medicine. Currently, however, there is not enough understood about the relationships between the genetic makeup of specific people and the genetic characteristics of specific diseases to, in most cases, safely make use this information in actual medical practice.The goal of our research is to learn more about these relationships. Most of what is known in this area relates to individuals genes and individual diseases. Much more can be learned however from studying interactions among groups of genes and diseases. Our approach models such interactions in the context of networks connecting genes and diseases, and examining significant interactions among multiple genes and disease. The networks we use are bipartite graphs, where there are two disjoint sets of nodes (typically genes and diseases) and relationships between individuals in one group and individuals in the other. We have developed an algorithm which finds all maximal bicliques in such networks. A biclique is a subnetwork where all nodes in one node set (say, genes) are related to all nodes in the other (say, diseases) The networks typically have thousands, even tens of thousands of node and hundreds of thousands of relationships. Adding to the challenge, the number of bicliques in a network, in the worst case, can grow exponentially with the number of nodes. Despite this, we have found the problem to be tractable in many realistic cases and we have successfully run our algorithm on problems of realistic size. We will present results showing that characteristics that make some networks tractable and also characteristics of bicliques found in these networks.
Mapping the Atmosphere of Mars, Season by Season
Tuesday, February 23, 2016
Speaker: Robert Novak, CFC, Ph.D., Physics Department
The composition of Mars' atmosphere is obtained by accumulating data using telescopes in Hawaii. The telescopic image of Mars is focused on the entrance slit of an infrared spectrograph attached to the telescope. The pointing position of the telescope and the settings on the spectrograph are adjusted by using computers at remote locations (such as Cornelia Hall). The images produced are analyzed to retrieve the molecular composition of Mars' atmosphere for a specific latitude/longitude point on Mars. With these values, a two-dimensional map is constructed. The talk described how computers are used in each step of the process and give examples of maps that have been constructed for water and ozone on Mars.
New Directions in the Use of Prosodic Features for Authorship Attribution of Historical Texts
Thursday, December 3, 2015
Speaker: Lubomir Ivanov, Ph.D., Computer Science Department
This seminar explored the use of certain prosodic features – lexical stress and alliteration – for authorship attribution of historical texts. Recent results were presented, which demonstrate that lexical stress may play an important role in improving the accuracy of attribution. The results also seem to suggest that some authors are more aware of the importance of prosodic features, and seem to employ them more effectively to strengthen the emotional impact of their work. Finally, the session outlined some future directions of research in the use of prosodic features for authorship attribution. The work presented in this talk is one aspect of a large and complex project on attributing mid- to late 18th century American and English literary works of political and historical significance.
The project, led by Dr. Smiljana Petrovic and Mr. Gary Berton, involves an interdisciplinary collaboration among faculty and students from the Computer Science and English Departments as well as the Thomas Paine National Historic Association.
Data Mining for Authorship Attribution
See slides from presentation
November 5, 2015
Speaker: Smiljana Petrovic, Ph.D., Computer Science Department
Data a mining is used extensively in authorship attribution of both modern and historical writings. This research concentrates on 18th century political writing, with a special focus on the work of one of the most influential and controversial figures of the time, Thomas Paine. Anonymously written essays were common at that time. Accurately determining the author helps historians to better understand the political, ideological, socio-economical and philosophical milieu of the period. In view of our results, some previously unattributed work was attributed to Paine, while some widely accepted attributions are brought into question.
The attribution problem here is approached as a classification task: a number of common and domain-specific stylometric features are extracted from undisputed writings of over 20 authors and analyzed using different machine learning methods to create a stylistic fingerprint for each considered author. A disputed work is then attributed to the author with the most similar style. A novel methodology that selects and combines base classifiers outperforms any individual classifier.
A number of professors and students from different departments at Iona are working on this project: Gary Berton, Sean Campbell, Dr. Scott Cleary, Dr. Lubomir Ivanov, Dr. Smiljana Petrovic and Dr. Robert Schiaffino. This presentation is the first of two talks dedicated to the Thomas Paine Authorship Attribution Project.
Semi-Quantum Cryptography: Security Proofs and New Directions
October 15, 2015
Speaker: Walter Krawec, Ph.D., Computer Science Department
Quantum key distribution (QKD) protocols allow two users Alice (A) and Bob (B) to establish a shared secret key, secure against even an all powerful adversary Eve (E). Semi-Quantum key distribution protocols (SQKD), introduced in 2007 by Boyer et al., are designed with the same goal in mind; however now, instead of requiring both A and B to support quantum operations, only Alice is required to do so while Bob is "classical" in nature. These protocols are interesting theoretically as they attempt to answer the question "how quantum does a protocol need to be in order to gain an advantage over a classical one?" They are also interesting practically as the classical user requires less complicated hardware. However, due to their reliance on a two-way quantum communication channel (and thus an attacker's ability to interact twice with the sent qubit), their security analysis is difficult. This talk will introduce QKD protocols and SQKD protocols. We will then discuss our recent proofs of unconditional security and the mathematical techniques employed. Finally we will discuss future possible research in this field. All required notation and concepts concerning quantum computation (e.g., qubits, measurements, operators, etc.) will be introduced in the talk.