Innovative Projects Realized

Suboptimal Decision-Making in Nectar-Drinking Bats – Cognitive Neuroscience

Robotics navigations in indoor and structured environment.

I will be working on the current project going on which involves studying the Robotics navigations in indoor and structured environment. Apart from that, we will be also looking into the SLAM problem in a humanoid robot (NAO). We are studying two approaches: (i) integration of SLAM with augmented reality, and (ii) SLAM via bio-inspired techniques. Both approached are expected to reduce the computational efforts compared to mathematical model-based SLAM and should also be useful in dynamic environments.
Autonomous navigation is regarded as an essential attribute of intelligent robots (mobile platform or humanoids). However, operating in human world is as important. More and more robots are finding their way to human environments such as factories, universities, hospitals, homes, etc. Perception of different environment and learning specialized tasks within that environment present host of challenging problems. TO BE CONT’D

Development of a biosensor library based on inexpensive printed circuit board technology

The first step is reviewing what has already been done in the lab and around the globe. Then we will Identify the biomarkers to target and techniques that can be used to do so. Next is to design a device in such a manner that the desired interaction is happening for efficient sensing of the target. Coordination with other students working on integration of the device will be maintained in order to ensure that sensor designs can be interfaced to the overall instrument. Then we will fabricate the sensor using machinery in the Micro Instrumentation lab and test the device against real samples. Depending upon its performance during testing phase, there will be some design changes to make it better. Some iterations will be done to come up with best design. Integrating this with the existing biosensor library will ensure possibility of multiplexing the tests. TO BE CONT’D

Extending Automated Segmentation methods for Body Composition Analysis

Body composition, i.e., the proportion of fat and muscle tissues in the human body is related to the risk factors associated with a host of medical conditions. The muscle and fat tissues are target locations for the water- and fat-soluble drugs respectively used for cancer treatment. Consequently, the proportions of these tissues are believed to determine
the chemotherapy toxicity and efficacy. Therefore, the estimation of muscle and fat tissue proportions is an important task in research studies related to cancer prognosis and treatment.
I want to introduce a fully automatic framework for the segmentation of muscle and fat tissues from CT images to estimate body composition. TO BE CONT’D

Model-Aided Performance Analysis from System Traces

System performance can be analyzed by measuring its operation, and by studying a performance model. Each has advantages: measurements have fidelity to the actual system, while models have predictive power. This work will join the two approaches, by creating a model from data collected from traces. If successful, this model will help Ciena to understand performance issues, and to maintain or improve performance as the system evolves.

Cannabis, Mindfulness and Yoga

Cannabis-enhanced yoga known as “Ganja Yoga” is an ancient practice that is burgeoning in North America, providing opportunities for both the cannabis industry and the yoga industry. While there is an accumulation of anecdotal reports regarding Ganja Yoga, there is yet to be research conducted on this topic. Delineating the prevalence, motives, mechanisms, risks, and benefits of Ganja Yoga can inform health practitioners, the yoga industry, and the cannabis industry as this practice continues to proliferate amidst permissive changes to the legal landscape. This study will utilize 3 phases including a survey, observational study, and an administration study. All phases will investigate cannabis as a mechanism to decrease barriers to accessing yoga, effects of the practice on health and well-being, and effects on mindfulness.

Medical geneticists’ discussion of psychiatric risks during diagnosis of 22q11.2 deletion syndrome

22q11.2 deletion syndrome (22qDS) affects 1/4,000 newborns. People with this condition can have various medical problems. Approximately 30% develop psychiatric illness (e.g bipolar disorder or schizophrenia). A recent study explored parents’ experience of receiving a diagnosis of 22qDS for their child. Families identified an unmet need for information from their healthcare providers about the psychiatric features of 22qDS, and indicated that risk for psychiatric illness was a major source of anxiety, compared to the other features of the syndrome. No studies have ever asked medical geneticists about how they approach telling families about the features of 22qDS. The purpose of our study is to find out if and when medical geneticists discuss with families different features of 22qDS, especially psychiatric risks. This project falls directly under several aspects of the mandate of BCMHAS. It explores the practices of physicians related to psychiatric disorders and helps the development of interventions.

Understanding Systematic and Firm-Specific Components in Credit Risk

Credit risk—the potential that a borrower will fail to make required payments—is the oldest risk in our economy. It may arise in a number of circumstances, for example, a consumer failing to make the minimum payment due on a credit card or a company defaulting on its debt. Our main goal in this summer project is to quantify both systematic—vulnerability to events which affect the whole economy—and firm-specific components in credit risk, and the interaction between them. During a 12-week period, the student is expected to propose a model that allows for both systematic and firm-specific risks, develop an estimation methodology for such a model, along pricing methods for credit-sensitive derivatives. He is also expected to perform the econometric estimation of the model using credit default swaps and give economic interpretations based on the model.

Extending and Deploying the Social CheatSheet Plugin

This project will expose the intern to key design, implementation, and evaluation research activities in human-computer interaction (HCI). We have recently started exploring the concept of social curation of software help content by developing a novel web-based platform, Social CheatSheet, that overlays relevant community-curated instructions and multi-step tutorials atop any web application and offers an easy curation interface for adding and editing content. In the proposed research, we will be extending the features of the Social CheatSheet platform to design user analytics and more intelligent content recommendation. Our goal is to contribute stronger insights into how users actually contribute and consume content related to software learning and help. For example, we will be able to investigate how participation rates compare across different domain-specific communities (e.g., users of spreadsheets vs. 3D modelling vs. health applications).

Complete classification of Littlewood cyclotomic polynomials

The objective of this proposal is to study Littlewood cyclotomic polynomials of odd degree. In algebra, the cyclotomic polynomial is one such that has all its roots on the unit circle. Since all the coefficients of Littlewood polynomials are -1 or +1, its associates a finite binary sequence with -1 or +1 entries. Therefore, their study is closely related to the study of finite binary sequences which is a basic object in the theory of information and communication. There is an extensive research in information and communication theory on studying the merit factors of finite binary sequences. Binary sequences with low autocorrelation coefficients are the most easily distinguishable signals and they are of interest in radar, sonar, and communication systems. It was observed that all Littlewood cyclotomic polynomials have very simple factorization as a product of irreducible Littlewood cyclotomic polynomials. TO BE CONT’D

Metabolic networks and applications to M. tuberculosis

The project involves the modeling of the metabolism of TB. I developed an algorithmic pipeline called MetaMerge, which allowed me to reconcile differences in format, nomenclature, and annotation, between two models of TB metabolism. MONGOOSE, another doctoral project of mine, is a tool for analyzing metabolic network models in exact arithmetic, resulting in consistent, reproducible predictions, something that is currently impossible with any other tools since they rely on floating-point arithmetic.
The student’s role will consist of completing the integration of MetaMerge into the MONGOOSE pipeline, automating the reconciliation of two metabolic network models using machine learning and data mining techniques, and running the algorithm on the metabolic network models of other organisms. This will allow us to create a unified user interface for the two programs, making both programs more user-friendly and accessible, as well as to refine it to work on other organisms. TO BE CONT’D

Learning Generative Models of Images and Patterns

This Project is an continuation of our SIGGRAPH Asia 2017 paper on “Learning to Group Graphical Patterns”.
The paper introduced a novel deep learning approach for grouping discrete patterns common in graphical designs. The approach was based on a convolutional neural network architecture that learns a grouping measure defined over a pair of pattern elements. Motivated by perceptual grouping principles, the key feature of the network was the encoding of element shape, context, symmetries, and structural arrangements. These element properties are all jointly considered and appropriately weighted in the grouping measure. To better align the measure with the human perception of grouping, the network was trained on a large, human-annotated dataset of pattern groupings consisting of patterns at varying granularity levels, with rich element relations and varieties, tempered with noise and other data imperfections. TO BE CONT’D