ECE graduate student Toluwanimi (Toluwa) Odemuyiwa selected for a Microsoft Ph.D. fellowship
- One of 45 students from around the world selected for this, she seeks to make it more efficient for computers to run on sparse data
Her current research interest is in algorithmic, software, and hardware techniques for finding novel ways to handle sparse data and computations as the need for high performance computing continues to grow in the age of big data. She is particularly interested in the ways software and hardware can be co-designed to produce novel solutions.
"My current research focuses on making it more efficient for computers -- whether CPUs, GPUs, or new platforms --- to work with sparse data. Our world runs on data, and these days, we’re producing more data than ever before. A lot of this data is sparse - that is, it contains missing information, either intentionally or unintentionally. Traditional architectures were not designed for sparse data. Over the past several years, there have been several advances in improving the computation of sparse data, such as compressing the data and designing smarter algorithms to make sparse computations much faster."
Odemuyiwa's goal is to co-design hardware and software to improve computation involving sparse data. "I’m working to understand the ways our computers fundamentally need to change to better support sparse computations, while also re-designing some of the sparse algorithms to fit these new computer designs. This in turn will make it more energy-efficient and faster to process all kinds of data our world runs on today."
A major way to save energy with computer data processing would be to find a way to more efficiently access data from main memory, as that takes up a lot of energy. Odemuyiwa is currently collaborating with researchers at NVIDIA, UIUC, and MIT on different partitioning strategies for sparse workloads in order to enable more efficient main memory accesses and better load balancing for computation.
Odemuyiwa is also exploring new ways to express graph algorithms in the language of sparse tensor algebra. Graph algorithms are the foundation of several applications, and fundamentally help us compute on anything that can be represented as a network of connections. Expressing graphs as sparse tensor algebra will enable us to leverage existing work in that area for faster graph algorithms.
She expects that in the future, Moore's law, which states that the number of transistors in a dense integrated circuit will double every two years, and therefore computers will become smaller and more able to process information at that rate, will no longer describe the state of computer engineering. The pace of development will pick up much more quickly.
Over the next few decades, Odemuyiwa expects computer architects and scientists to focus on two main areas of research. "One, identifying the key applications and problems for which our world needs more efficient computing (such as data analytics) and two, designing new software algorithms with the underlying hardware platform in mind and vice versa, as co-design is important now more than ever. Moreover, emerging systems will contain a diversity of nodes including CPUs, GPUs (graphic processing units), and special-function accelerators. We need to rise to the challenge of addressing networking, security, architecture design, and programming of these systems."
She feels incredibly supported and encouraged by her professors, mentors and colleagues at UC Davis and is glad her work will benefit both academia and industry. As she explains, "I am grateful to the electrical and computer engineering (ECE) department for nominating me for this fellowship. It is very humbling to know that my professors could place so much confidence in me. The fellowship provides me with an opportunity to be mentored by world-class researchers at Microsoft. Over the past couple of years since completing my undergraduate degree, I have come to recognize the importance of university-industry collaborations in identifying and solving practical problems. The ECE department at UC Davis has provided that kind of platform since I arrived; I feel fortunate that this scholarship will allow me to further expand it to Microsoft."
She further expands on what she most loves about UC Davis, saying that joining the ECE department was one of her best career decisions. "In the two years since I’ve been here, I’ve received immense support and encouragement not just from my advisor, but other professors within the department and in CS. The ECE-GSA has also provided an excellent community where colleagues have quickly become friends. The administrative staff are always on top of things and they make it easy for us to focus on our research and academic work. For research, one major advantage of being at Davis is the presence of a wide variety of cross-departmental (interdisciplinary) research groups that make it possible to see problems from different perspectives. I believe this kind of environment builds a unique type of researcher."
She highly praises her thesis advisor, electrical and computer engineering professor John Owens, for his insight, knowledge, and support. "Most importantly, I am proud to be a part of an amazing research group led by John Owens, where I’m consistently challenged and intrigued by the work of my highly talented colleagues. From connecting me with internship opportunities, to world-class researchers at various institutes, to constantly watching out for me and simply providing me feedback or suggestions on my work, I couldn’t have asked for a better advisor and mentor in John."
Odemuyiwa finds mentorship one of the strongest aspects of UC Davis' engineering academic programs. "Overall, I would say UC Davis is extremely strong in mentorship opportunities -- both in being mentored as a grad student, and in also mentoring others. I’ve had the privilege of being mentored by Prof. Stacey Combes through the GSoC program. In addition, the PROMISE Engineering Institute and the ADEPT programs provide me with opportunities to network with peers in other departments and learn more about career options post-graduation."