HDSI Names Faculty Fellows, Class of 2020

Photo of HDSI Lapel Buttons

by Trista Sobeck

Rajesh Gupta, HDSI Founding Director, has announced the appointment of the following HDSI Faculty Council members as HDSI Faculty Fellows, Class of 2020. This two-year appointment comes in recognition of their contributions to the Halıcıoğlu Data Science Institute and its mission and programs.

These Fellows have shown their willingness and passion for the emerging community of HDSI affiliates through such projects as streamlining cyberinfrastructure resources and managing senior design projects, while keeping HDSI leadership engaged and informed.

Congratulations to the following faculty members:

Headshot of Ilkay Altintas

Ilkay Altintas, Founding HDSI Fellow

Ilkay Altintas is a Founding Fellow of the Halıcıoğlu Data Science Institute (HDSI) as well as San Diego Supercomputer Center’s Chief Data Science Officer. In addition to her CDSO responsibilities at SDSC, an Organized Research Unit of UC San Diego, Altintas is an associate research scientist and the founding director of the Workflows for Data Science (WoRDS) Center of Excellence at SDSC. WoRDS specializes in the development of methods, cyberinfrastructure and workflows for computational data science and its translation to practical applications. The WIFIRE Lab she founded and directs is focused on artificial intelligence methods for an all-hazards knowledge cyberinfrastructure, becoming a management layer from the data collection to modeling efforts, and has achieved significant success in helping to manage wildfires.

Headshot of Arun Kumar

Arun Kumar

Arun is a member of the Database Lab and Center for Networked Systems and an affiliate member of the Artificial Intelligence group. His primary research interests are in data management and systems for machine learning-based data analytics. Systems and ideas based on his research have been released as part of the Apache MADlib open-source library, shipped as part of products from Cloudera, IBM, Oracle, and Pivotal, and used internally by Facebook, Google, LogicBlox, Microsoft, and other companies. His current research focuses on simplifying and accelerating the processes of data preparation, model selection, and model deployment.

Tara Javidi

Tara founded and co-directs the Center for Machine-Integrated Computing and Security at UC San Diego. Among her projects is DetecDrone Target Search Technology. Her research focus is on stochastic analysis, design, and control of information collection, processing, and transfer in modern communication and networked systems. Her work also includes the applications of microeconomic theory and organizational science to the design of wireless networks.

Brad Voytek, Founding HDSI Fellow

Bradley Voytek is an Associate Professor in the Department of Cognitive Science, the Halıcıoğlu Data Science Institute, and the Neurosciences Graduate Program at UC San Diego. He is an Alfred P. Sloan Neuroscience Research Fellow and National Academies Kavli Fellow, as well as a founding faculty member of the UC San Diego Halıcıoğlu Data Science Institute and its undergraduate Data Science program, where he serves as Vice-chair. His research program uses neural modeling and simulation, along with large-scale data mining and machine learning techniques, to understand the physiological basis of human cognition and age-related cognitive decline.

For questions regarding this article and other HDSI information, please contact HDSIComm@ucsd.edu.

Congratulations 2021 HDSI Scholarship Recipients

graphic logo stating HDSI Scholarship Program

by Trista Sobeck

Congratulations to 28 Halıcıoğlu Data Science Undergraduate scholarship recipients who have shown they are motivated to find data-driven solutions to real-world problems. In HDSI’s mission of supporting multidisciplinary, student-led projects, our 2021 recipients show promise as future Data Science leaders.

The program was open to all UC San Diego undergraduate students, with priority given to students in the Data Science major or minor. Here are some highlights from this year’s recipients:

Camille Dunning was inspired by her recent neurobiology coursework and involvement in IEEE Project Drive. Her interest in reinforcement learning was piqued. “I am also focusing on signal processing (ECG waveforms and economic data in the past) and Natural Language Processing,” she explains. She recently founded a startup that is working to improve current NLP practices and actualize a fully functioning and smart product.

Michael Garcia-Perez  is excited to use data in order to help capture regional COVID trends by applying natural language processing techniques. “The data that I specifically work with consists of various Twitter data and Google Trends. The sentiment provided from tweets will allow me to identify trends in COVID rates which will help me answer my research,” he says. He is observing regional COVID rates so that he can discover trends that might be preventable. 

Amelia Kawasaki originally came to data science from cyber security. After seeing how data scientists were able to process and train models on network and file data to improve cyber security on local networks, she became inspired. “It was the first time that I was able to see how math and computer science skills could directly protect people. Data science is the perfect union of math and computer science that I didn’t know I wanted,” she explains. The scholarship will enable her to focus on her project over the summer so she can begin her career in research and enter grad school.

Bailey Man is interested in using data science for projects spanning from GIS data, and website user purchases, to COVID particles and recently, dolphin cognition. This project is unique because it attempts to gain insight into the conversations and sounds made dolphins and provides the architecture for much further research. “The types of problems I envision myself solving are ones that utilize massive amounts of data, not simply for the sake of its size, but also because the scale is only recently becoming possible,” he says.

Sam Schickler has been interested in data since the 7th grade. “I want to use data to help people understand our world better. Whether that is in neuroscience, where I am currently helping to develop software and algorithms to help scientists understand the brain, or in economics or in political science,” he says. One of Sam’s goals is to help enable the use and development of CIDAN (Calcium Image Data Analysis) to process brain images.

Sirui Tao would like to explore the possibility of using AI to facilitate better design interaction, optimize the design process. “The scholarship helps introduce me to Prof. Judith Fan’s Cognitive Tools lab, which gives me lots of opportunities to learn from other researchers who are actively approaching the problems in various other ways,” he says. He is working on a project that will showcase how AI can be used in a highly creative field.

Zirui Wang who also goes by Colin, has been an EDM producer for a few years and started to wonder if artificial intelligence could be used to put the music together. “Starting from here, I began to read data science books and fell in love with it,” he says. Colin is also working on a proposed project that aims to extract useful information from any single audio/music spectrogram and uses a CNN-based stacked GAN to generate similar music based on that.

Watch some of last year’s HDSI Undergraduate Scholarship presentations here.

For questions regarding this article and other HDSI information, please contact HDSIComm@ucsd.edu.

 

The Importance of Inquiry Across Disciplines: Top Deep Math Conference Features Halıcıoğlu Data Science Professors and Faculty Members

deep math conference logo

by Trista Sobeck

On Nov 5, 2020, a virtual Deep Math conference took place featuring several HDSI faculty members. The Conference on the “Mathematical Theory of Deep Neural Networks” has a mission—especially this year—to bring together researchers from several fields. Much like HDSI’s interdisciplinary ethos, this conference focused on what deep learning means through different perspectives.

However, Assistant Professor Mikio Aoi with joint appointments at UC San Diego’s Halıcıoğlu Data Science Institute and Biological Sciences & Neurobiology would like to flip that idea on its head.

Aoi has been with the conference planning board since its inception—three years ago–and in fact, is one of the core co-organizers. He explains deep learning as a type of Rorschach test.

“[P]eople see in deep learning the signatures of their own perspectives. Our view is that if people from different backgrounds could be brought together to think about the problems of understanding deep learning through different lenses, then perhaps the cross-pollination of ideas could accelerate our understanding,” says Aoi.

This conference also allows for the freedom of those interested in expanding their specific fields of research. According to Mikhail (Misha) Belkin, Professor at the Halıcıoğlu Data Science Institute and one of eight speakers invited to the conference, deep learning has been one of the most recent developments in perhaps all of science.

“We do not yet have a strong mathematical theory for deep learning. Indeed, the practice has shown significant gaps in our mathematical understanding of learning phenomena.”  He says a fundamental mathematical theory is key for conceptual understanding and improving existing methods and building new algorithms.

To put it plainly, we need to look at things differently – at all times if we are going to create new ideas.

According to Gal Mishne, Assistant Professor at Halıcıoğlu Data Science Institute and one of the co-organizers of the conference this year, there is still much unknown about why deep learning is so successful. “The Deep Math conference brings together researchers from Mathematics, Physics, Machine Learning, Signal Processing, Neuroscience and more who otherwise might not necessarily interact, to discuss and explore different perspectives underlying the theory of deep learning,” she says. Furthermore, the conference format encourages debate among the invited speakers beyond their individual presentations.

What makes this conference so unique? Many things. But, according to Aoi, to his knowledge, this is the only one focused on the discussion and promotion of theories of deep learning across disciplines. “There have been many ad hoc events, many of which were either attached to larger conferences or were focused on specific disciplinary perspectives.”

Belkin concurs and says that deep learning has changed everything in the last 10 years and feels fortunate to be able to participate in developments “[T]his is an exciting time, certainly the most exciting time of my career, perhaps even in the whole history of the whole subject of machine learning,” he says.

In what first was developed as a casual conversation about deep learning while he was a postdoc at Princeton, Aoi recounts that he was part of a journal club for theories of deep learning.

“There were few papers at the time, and we had many questions we couldn’t answer. We realized that many of the people we would like to hear from were at the Institute for Advanced Study (IAS) just down the road.

This independent postdoctoral research center for theoretical research and intellectual inquiry has served as the academic home of internationally preeminent scholars, including Albert Einstein and J. Robert Oppenheimer. “This gave us the idea of a mini-conference of local speakers, but as local interest grew, we realized that there was enough demand for a larger event with multiple speakers.”

And here we are today where some of the top researchers on mathematical theory of deep learning and related topics gather. “It is a great opportunity to communicate recent progress to an audience like that and learn about new developments in the area,” says Belkin.

Mishne, the first faculty member hired at HDSI says she has a goal to help define and develop the new field of data science. Organizing such conferences bringing together researchers from multiple disciplines is just one step. She wants to answer the ongoing question, ‘what is the core knowledge and skills that a data scientist needs?’

“I hope my students learn that being a data scientist requires [skills] beyond just good analytical and coding skills,” says Mishne. “It requires curiosity–a drive to understand and explain information hidden in data, as well as communication – being able to present and explain your findings in an effective manner to different audiences.”

The next annual Deep Math Conference is scheduled for November 2021.

For questions regarding this article and other HDSI information, please contact HDSIComm@ucsd.edu.

 

Pivoting from laboratories to nature: Outdoor classes allow for in-person education to continue

UC San Diego outdoor classroom tent

by Trista Sobeck

UC San Diego is innovative and ahead of the curve. With the university’s goal to “Return to Learn”, the Halıcıoğlu Data Science Institute is again at the forefront of thinking outside the box – or lab in this case.

Because the Covid-19 pandemic hit the world – and the United States—hard in 2020, learners of all ages were forced to learn from the safety of their homes and their laptops. This presents a huge challenge for not just the students, but for the educators too.

With many classes being hands-on and collaborative, some educators were forced to pivot from their typical learning environment such as indoor labs to embracing the out-of-doors, in almost bespoke locations.

Jack Silberman portrait

Jack Silberman Ph.D., HDSI Lecturer and Faculty at the Contextual Robotics Institute, took DSC 190A00 – Introduction to Robotics Perception and Navigation class (in conjunction with the Jacobs School of Engineering’s ECE MAE 148 – Introduction to Autonomous Vehicles) outside and out of the lab. And, unbeknownst to anyone at the time, it was largely a success.

Students continue to follow all safety protocols and of course, wear masks. The only issue Dr. Silberman has had, and even experiences firsthand, is foggy glasses and dressing warmly. “At this point, we all know it is either using a mask or not coming to campus,” he says. “Besides that, we follow the protocols to keep social distance, sanitize our hands often, and no face touching.”

This class is very popular with many students because of the hands-on industry-relevant application. “Students feel it is hands-on from the first day of class,” says Dr. Silberman. “They are getting a tactile experience. These students are interested in field robotics and for several of them, this is the first class where they can experience robotics and with a physical robot,” he explains.

Therefore, it is helping students stay socially active while reducing risks and taking the class they had looked forward to, including working in teams. Dr. Silberman is even getting thank you notes and emails from students who want to convey their excitement that his class is still in session:

“Hello Professor Silberman,

It looks like I was selected, and I am authorized to enroll in the course! I just wanted to say that I really want to take in-person labs next quarter, I’m going a bit crazy sitting at home all day. I think it would improve my overall mood and possibly even my performance in other courses….”

In this strange time, students and faculty must find collaboration and develop a positive attitude in their work. Whether it is enabling hands-on robotics classes or through socially distant, face-to-face communication while managing risks, it’s dire that we all remain hopeful, engaged, have positive attitudes. Of course, we need to continue to learn.

“This year is all about getting everyone comfortable learning while managing risks, stress, and uncertainties. Eventually, students will realize that this experience made them more confident in solving challenges and trusting that they have the grit to succeed in what comes next after UCSD,” Dr. Silberman explains. And because of his inventive thinking, fundamental learning won’t have to stop.

For questions regarding this article and other HDSI information, please contact HDSIComm@ucsd.edu.

 

 

Data Science Faculty Featured at 2020 NeurIPS Annual Meeting

logo conference on neural information processing systems

by Bobby Gordon

Several faculty from the Halıcıoğlu Data Science Institute at UC San Diego are featured presenters at the 2020 NeurIPS Annual Meeting this week.  Featuring peer-reviewed research and talks by industry and field leaders, “the purpose of the Neural Information Processing Systems (NeuIPS) annual meeting is to foster the exchange of research on neural information processing systems in their biological, technological, mathematical, and theoretical aspects,” according to this year’s Annual Meeting site.

Mikhail (Misha) Belkin, Professor with the Halıcıoğlu Data Science Institute, is leading a presentation based on his ongoing work with his Ph.D. students (Chaoyue Liu & Libin Zhu).  “A recent discovery about neural networks is that certain very large neural networks are essentially linear functions of parameters,” says Belkin.  He adds, “This is very surprising since the structure of neural networks is highly non-linear and it is not clear why this ‘transition to linearity’ should occur for wide networks. In our work we provide a new perspective on this phenomenon, showing why it happens and demonstrating that it is not a general property of big systems, but is specific to certain architectures.”  The team is asking key questions related to neural networks and identifying whether they have advantages of better understood and more traditional kernel methods; and if yes, identifying whether they can analyze those advantages.  Belkin plans to build upon these findings in their on-going research, and plans to incorporate these into his HDSI Data Science courses as well.

Rose Yu, Assistant Professor with Computer Science and Engineering and the Halıcıoğlu Data Science Institute, is also a featured presenter this week.  Her work, supported by an Army Research Office grant in partnership with colleagues from Northeastern University, focuses on building “an intelligent machine that can reason about what objects are, how they move, and what happens when they are missing in videos.  We also want to do this in a completely self-supervised fashion without labeled training data,” according to Yu.  She adds, “A 5-month-old child can understand that objects continue to exist even when they are unseen, a phenomenon known as ‘object permanence.’   However, current deep learning methods cannot reason about the objects when they are missing in videos.  Our method can simultaneously perform object decomposition, latent space disentangling, missing data imputation, and video forecasting.”  Yu has already incorporated this work into her curriculum at UC San Diego, and she’s working with HDSI colleagues to apply this model to COVID-19 forecasting.

Julian McAuley, Associate Professor with Computer Science and Engineering and the Halıcıoğlu Data Science Institute, is part of a team project originally supported by Microsoft Research Asia; proposing a dynamic acceleration method for large language models.  Their unique approach “terminates the forward pass of a neural network early using ‘patience’ as a signal,” according to Canwen Xu, Ph.D. student with Computer Science and Engineering at UC San Diego.  “Our idea came from neural network training.  We found the similarity between training and inference of a neural network.  This research highlights the ‘overthinking’ problem in neural networks and opens a new door for faster inference, and more efficient neural networks,” continues Xu.

Other Halıcıoğlu Data Science Institute faculty featured in this week’s schedule include Yian Ma Assistant Professor, Halıcıoğlu Data Science Institute, and Henrik Christensen, Director of the Contextual Robotics Institute; also with Computer Science and Engineering and the Halıcıoğlu Data Science Institute at UC San Diego.

Learn more about the 2020 NeurIPS Annual Meeting schedule here.

For questions regarding this article and other HDSI information, please contact HDSIComm@ucsd.edu.