If you accept a Microsoft Research Dissertation Grant, you may receive another fellowship from another company or institution during the same academic period.
Experience shows us that people with disabilities can positively impact interaction design for everyone. However, publishers of interaction design rubrics—such as Human-Centered Design—have tended to focus on supporting the design process for people with disabilities, rather than by them. My research focuses on developing an inclusive toolkit that augments current Human-Centered Design activities to be accessible to people with disabilities. Drawing from this toolkit, I will offer new ways to connect disability with design, all based on the life experiences of people with disabilities.
The work of community engagement performed by public officials in local government provides valuable opportunities for city residents to participate in governance. Technology stands to play an increasingly important role in mediating community engagement; however, the practices and relationships that constitute community engagement are currently understudied in human-computer interaction HCI.
Of particular importance is the role that trust plays in the success of community engagements—either establishing trust, or more frequently, overcoming distrust between public officials and city residents.
To address this challenge, my research seeks to understand how trust could inform the design of technology to support the work of community engagement performed by public officials in local government. My research will culminate in a design framework that will inform development of technology for trust-based community engagement.
Augmented listening technologies, such as hearing aids, smart headphones, and audio augmented- reality platforms, promise to enhance human hearing by processing the sound we hear to reduce unwanted noise and improve understanding. State-of-the-art listening devices perform poorly, however, in noisy environments that have many competing sound sources. Large microphone arrays with dozens or hundreds of sensors could allow listening devices to separate, process, and enhance multiple sound sources in real time while sounding natural to the user.
I am also developing first-of-their-kind wearable microphone array prototypes and data sets to help other researchers develop ambitious new augmented listening algorithms and applications.
Machine learning is increasingly being used for decision support in critical settings, where predictions have potentially grave implications over human lives. Examples of such applications include child welfare, criminal justice, and healthcare. In these settings, the characteristics of available data and of deployment contexts give rise to challenges that have not been sufficiently addressed in the machine learning literature, including the presence of selective labels, unobservables, and the effects of omitted payoff bias.
When left unaddressed, these challenges may lead to systemic biases, self-fulfilling prophecies, and loss of human trust in the systems. My research is focused on quantifying the performance and fairness risks of algorithmic learning in these settings, and on reducing these risks by developing novel algorithms. Providing Context for Capture-Time Decisions. As cameras become smarter and more pervasive, more people want to learn to be better content creators.
People are willing to invest in expensive cameras as a medium for their artistic expression, but few have easy ways to improve their skills. Inspired by critique sessions common in in-person art practice classes, my dissertation research focuses on designing new interfaces and interactions that help people become better photo takers. Using contextual in-camera feedback, users can capture photos and videos in a way that is more informed and intentional, while still allowing for their aesthetic and creative decisions.
Highly interactive modeling methods and audio enhancement algorithms underlie the operation of modern acoustic systems. The capability of a system to produce lifelike acoustic experiences significantly depends on the accuracy and computational efficiency of the modeling and audio processing algorithms employed. Accordingly, my research has focused on the development of methods and algorithms that accurately model highly reverberant acoustic systems and process acoustic signals using as few parameters as possible.
Such accurate yet computationally efficient modeling and processing algorithms are of essential interest in a wide variety of applications ranging from virtual acoustics to healthcare. My main contribution is the development of algorithms, which rely on orthonormal basis functions and time-frequency representation of an acoustic system, that provide high accuracy over a wide range of frequencies in real-time.
As an early demonstration, I propose an efficient solution to adaptive feedback cancellation problems. Major advances in computer vision and mobile technologies have set the stage for widespread deployment of connected cameras, spurring increased concerns about privacy and security.
Moving forward, I aim to leverage this framework to build low-power privacy-preserving computational cameras with camera-level implementations of learned encoding functions.
Deploying AI systems safely in the real world is challenging. The rich and complex nature of the open world makes it difficult for machines trained on limited data to adapt and generalize well. The errors that can result from an imperfect model can be extremely costly e. My research focuses on using human feedback to help reinforcement learning agents better adapt to the real world, leading to safer deployment of these systems.
This involves developing robust models that can accurately predict uncertainty in the world, use different forms of human input to learn, and adapt quickly in real-time to new changes in the environment. Developing such systems that learn from humans intelligently will move us closer towards more generalizable robots that perform a variety of tasks in such applications as assistive robotics, healthcare, and disaster response.
There has been a renewed focus on dialog systems, including non-task driven conversational agents i. Dialog is a challenging problem since it spans multiple conversational turns. To further complicate the problem, there are many contextual cues and valid possible utterances. We propose that dialog is fundamentally a multiscale process, given that context is carried from previous utterances in the conversation.
Neural dialog models, which are based on recurrent neural network RNN encoder-decoder sequence-to-sequence models, lack the ability to create temporal and stylistic coherence in conversations.
My thesis focuses on novel hierarchical approaches to improve the responses of neural chatbots. To that end, modern network devices offer programming interfaces for fine-grained specification of what information to maintain across packets, and how to process packets based on it.
My thesis focuses on designing programming platforms that facilitate the use of programmable network devices for large-scale and real-time network monitoring and control. More specifically, these platforms consist of i domain-specific languages that are expressive enough for high-level specification of policies for end-to-end network transport, network-wide state-aware monitoring and control, and path-based network monitoring, and ii compilers that use efficient intermediate data structures to automatically distribute and implement these specifications on programmable network devices.
I aim to develop methods to help users of machine learning models increase both the trust in and understanding of their models. My dissertation is in the two fields of interpretability and causal inference.
The two fields, seemingly disparate, actually share the common goals of revealing and adjusting for biases that can arise when building machine learning models. In causal inference, I have worked on methods that use machine learning to more flexibly estimate treatment effects from observational data. To complete my dissertation, I plan to probe the definition of interpretability — still a subject of debate in machine learning — by conducting a large-scale comparison of different models claimed to be interpretable and augment this quantitative evaluation with human subject experiments using domain experts.
Ebuka Arinze Johns Hopkins University. Nanoengineering for Tunable Energy-Efficient Optoelectronics. Colloidal nanomaterials, such as semiconductor quantum dots, are of interest for various optoelectronic applications due to their size-tunable optical properties, distinctive electronic structure, and low-cost fabrication.
Color-tuned and semi-transparent photovoltaics, devices with controlled and tunable reflection and transmission spectra, are of significant interest due to their potential applications in building-integrated photovoltaics, vehicular heat and power management, and multijunction photovoltaics.
My project focuses on using nanoengineering techniques, including multi-objective optimization algorithms, plasmonic nanoparticle enhancements, and hybrid-materials-based surface modifications, to design and build colloidal quantum dot-based devices with controlled optical and electrical properties for the next generation of inexpensive and ubiquitous light harvesting, detection, and emission technologies.
These algorithms allow us to specify data collection tasks, e. To reduce the amount of data needed for each task, and since models of underwater dynamics are computationally expensive, we use model-based reinforcement learning techniques where the models are data-driven. A problem with these approaches is that, even if they are data efficient, collecting new data is expensive.
Adopting cloud services to reduce operational, maintenance and storage costs, is becoming increasingly common. However, outsourcing data and computations, is opening up new challenges in terms of integrity and privacy of the data and the computations on them. Along with such important security and privacy concerns, availability, and scalability are major factors in such settings. My thesis addresses various problems in this space of secure storage and computation outsourcing.
In summary, the main contributions of my thesis are the following. The beginning of a new era in safe assistive robotics will occur when people with disabilities and seniors let intelligent software control a mobile robotic manipulator to safely reposition their body and limbs. Our goal is to explore the intersection between providing physical care and robotics, and how it is possible to translate safe patient handling and mobility guidelines into smart human-robotic interaction HRI algorithms.
For a mobile manipulator with knowledge-managed algorithms. Our efforts seek to standardize protocols and regulations for how artificial intelligence agents related to physical HRI can achieve body and limb repositioning tasks. As assistive robotics become more mainstream, these best practices can improve safety in direct physical care in the process of repositioning the human body with a mobile robotic arm.
My research primarily focuses on exploring how machine learning can help improve real world decision making in domains such as health care and criminal justice. To this end, my thesis addresses various challenges involved in developing and evaluating interpretable machine learning frameworks which can complement and provide insights into human decision making.
More specifically, my thesis focuses on the following diverse yet related research directions: The main contribution of my thesis is to address these problems under realistic assumptions which hold in real world decision making such as presence of unmeasured confounders and limited availability of labeled data.
My study examines the implementation of the health information system HIS in Mozambique and the roletechnologies play in educating health professionals for better delivery of care. Through a comprehensive examination of the HIS, from development to roll-out, I analyze the relationship between colonial and post colonial governmental top-down policies and compare them to the on-the-ground reality of using information and communications technology ICTs to provide health education given social, economic, and political realities in Mozambique.
Part of the problem with studies of technologies in poor parts of the world is that they are often conducted by highly educated researchers and are conducted in English. However, majority of the population in poor nations does not speak English. Such studies become irrelevant to the life experiences of those being studied. I will disseminate findings from this study in Portuguese and English through talks and publications in U. Students may be nominated for only one grant period.
Directors of Graduate Programs should submit the nomination form , as well as supporting documentation as indicated on the form to Dr. Applications will be reviewed by a committee appointed by the Graduate School. Grant periods for are as follows: July 1 to December 31, , and January 1 to June 30, The deadline for receipt of nominations for the first grant period is May 16, , and for the second grant period is October 16, Awards will be announced by June and November , respectively.
For further information, please contact Dr. Financial Support Residency Priority Deadlines. Summary of Procedures 3. Summary of Requirements 3. Learning the language Postdoc Overtime: Graduate School Doctoral Dissertation Completion Grants Program Summary The Doctoral Dissertation Completion Grant program provides both funding and intensive mentoring to doctoral candidates who are within six months of completing their dissertations.
Eligibility This grant is for students who have demonstrated difficulties in completing the dissertation and would benefit from the intensive mentoring and the six months dedicated to writing that are provided by this program.
The Doctoral Dissertation Completion Grant program provides both funding and intensive mentoring to doctoral candidates who are within six months of completing their dissertations. It is designed to enable candidates to focus full time on the writing of their dissertations, improving the quality of the dissertation and shortening the time.
DOCTORAL DISSERTATION GRANT PROGRAM Application Guidelines Areas of Funding Grants of up to $5, are available to help support dissertation expenses of doctoral students in the United States and Canada whose studies have the potential for adding significantly to Doctoral Grant Guidelines.
Doctoral Dissertation Grant Application Faculty/Post-Doctoral Grant Program (Fahs-Beck Fellows) Grants of up to $20, are available to help support the research of faculty members or post-doctoral researchers affiliated with non-profit human service organizations in the United States and Canada. February 23, Fahs-Beck Doctoral Dissertation Grant. Application Deadline: April 1, The Fahs-Beck Fund provides grants of up to $5, to help support dissertation expenses of doctoral students in the United States and Canada whose studies have the potential for adding significantly to knowledge about problems in the .
Microsoft Research is funding a new academic program, the Microsoft Research Dissertation Grant, offering selected doctoral students doing computing research at U.S. and Canadian universities up to US $20, to fund their dissertation work. Program Description. Doctoral Dissertation Research Grants help Doctoral candidates who have been accepted to an accredited institution complete research and dissertations on housing or urban development issues.