“It was really incredible to see how all knowledge I’ve garnered from U of T gave me such a strong basis for working on these projects”
Lana El Sanyoura will start her final year of the undergraduate program in Computer and Cognitive Science next fall. Lana was a member of the Computer Science Student Union (CSSU) and helped organize 2017's Frosh orientation event for first-year computer science students. Last year, she was awarded Konrad Group's Women in Technology scholarship.
Computer Science Communications caught up with Lana to learn more about the the research she undertook this summer in Boston, MA before starting her Professional Experience Year (PEY) co-op program at Intel in Toronto.
Where did you undertake research this summer?
I did research at the MIT-IBM Watson AI Lab within IBM Research in Cambridge, Massachusetts, under the supervision of Professor David Cox. This joint collaboration between IBM and MIT is focused on advancing four research pillars: artificial intelligence [AI] algorithms, the physics of AI, the application of AI to industries and advancing shared prosperity through AI.
How did you apply for your internship?
An IBMer knew of my interest in Computer Science and Cognitive Science and submitted my resumé through a referral program. The interview process picked up from there and I joined the lab as an undergraduate research intern for the summer of 2018!
What interested you in this opportunity?
It was a combination of everything that I am passionate about. I'm studying Computer Science and Cognitive Science and have a great interest in the philosophy of mind, the applications of AI, and the overlap between these two domains.
Some of my favorite discussions at office hours used to include questioning what an AI system's behavior could possibly tell us about our own minds, and how we might take inspiration from our own cognition to make AI systems better.
By joining the MIT-IBM Watson AI Lab, I hoped the experience would surround me with cutting-edge AI research, allow me to explore more questions, garner more answers, and grow as a researcher, computer scientist and cognitive scientist. And I was not mistaken!
What was your project about?
Something that I really loved about my internship was the collaborative and supportive environment that allowed me to explore and work on different projects throughout my term.
When David and I met to decide on the project I would be working on, he asked me what I was passionate about, where my interests in AI lie, and what I might like to explore during this internship. Hearing about my interest and experience with Computer Science, Cognitive Science, and the biological plausibility of AI systems, he suggested a project involving PredNets.
A PredNet, developed by William Lotter, Gabriel Kreiman, and David Cox, is a neural network architecture for next-frame video prediction that takes inspiration from Predictive Coding in Neuroscience. This summer, I was working on making this architecture more scaleable and studying the impact of additional features on its performance.
Along with this project, I was doing research on neural network regularizers with Dima Krotov and have also published tutorials on using IBM's Deep Learning service to perform hyperparameter optimization and run the scaled-up PredNets.
I also had the opportunity to serve as a program committee member for the IEEE VIS2018 VISxAI workshop in Berlin, Germany! Neural network explainability interests me from a philosophical, ethical and technical standpoint and the VISxAI workshop will be showcasing some of the field's leading neural network explainables.
It was really incredible to see how all knowledge I've garnered from U of T, all the discussions I've had with my professors and friends, gave me such a strong basis for working on these projects, partaking in exceptional technical and philosophical discussions and collaborating on exciting research.
What was most challenging about this project? Did you overcome any setbacks?
A challenging aspect of working on optimizing the PredNet architecture was setting-up and validating different hypotheses. I overcame setbacks through diagnosing the causes of the errors, reading the literature, and brainstorming with my colleagues. It was really important to get feedback on the modifications we can make to our hypotheses and explore new experimental directions.
For example, I remember being in the snack room one time, having a really nice conversation with Hendrik Strobelt and Dylan Cashman about data privacy, when the discussion veered towards me telling them about a small roadblock I was facing in my project.
We started discussing the issue further, until Hendrik said: "Let's go get a whiteboard."
The three of us sat in a meeting room and hammered out the problem; we asked hard questions, proposed potential answers, and came out with a solution to the issue. Even though Hendrik and Dylan are on a different team, and are not working on my project, they invested their time in helping me, which was really great.
What's next for you this year?
The year is off to a great start! I just received the Joseph Wesley MacCallum Scholarship from the Senate of Victoria University [at U of T] and will be starting a 12 month PEY co-op back in Toronto with Intel.
I'll be on the Deep Learning Acceleration (DLA) team, working on implementing machine learning algorithms on FPGAs [field-programmable gate array].
On campus, I'm thrilled to be a panelist and orientation leader at this year's computer science freshman orientation, aka Frosh, and will be rejoining the Victoria College's intramural basketball team.
What are you looking forward to, when returning to U of T campus next fall?
When back on campus after PEY co-op, I'm looking forward to getting more involved with the Computer Science student community, taking more machine learning and Cognitive Science courses – and graduating!
This summer has been truly incredible. Major thanks to my colleagues, professors, friends and family. It's been a great time here in Boston and I've loved every second of it!